Ministral 8B API Pricing: Symmetric Token Rates for Edge & Cloud
Examine Ministral 8B API pricing ($0.15/M input, $0.15/M output), symmetric token rates, edge on-device inference, and sub-second classification speed.
Full specs, context window and API limits →How much does Ministral 8B cost per million tokens?
Ministral 8B costs $0.15 per million input tokens and $0.15 per million output tokens ($0.15/M blended at 3:1). Features unique symmetric pricing, 128K context, and ultra-compact edge deployment. Verified 2026-09-08.
Verified 2026-09-08 — source
How much does Ministral 8B cost per 1,000 requests?
Computed from generated token pricing. Each row assumes the listed input and output tokens per request; output is adjusted by this model's measured 0.93× verbosity factor.
| Request shape | Input tokens | Output tokens | Cost / 1,000 requests |
|---|---|---|---|
| Short | 100 | 50 | $0.0220 |
| Medium | 1,000 | 500 | $0.2198 |
| Long | 4,000 | 2,000 | $0.8790 |
Formula: ((input price × input tokens) + (output price × output tokens × verbosity factor)) ÷ 1,000,000 × 1,000. Assumptions: short 100/50, medium 1,000/500, long 4,000/2,000 input/output tokens per request. Verbosity run: 2026-06-16T20:31:30.728Z.
Three model-specific pricing decisions
Ministral 8B is evaluated as an edge-class, high-volume API row. Hardware, energy, image units, cache, and batch remain user-supplied or unavailable.
1. Requests-per-month ladder for routing, tagging, and extraction
| Workload | Input / output | 100K requests | 1M requests | 10M requests |
|---|---|---|---|---|
| Routing | 200 / 30 | $3.45 | $34.50 | $345.00 |
| Tagging | 500 / 80 | $8.70 | $87.00 | $870.00 |
| Short extraction | 1,000 / 250 | $18.75 | $187.50 | $1875.00 |
Formula: requests × token shape × listed $/M. Output expansion is not silently added.
2. Equal-input-rate audit against Mistral Small
| Fixed input / output | Ministral 8B | Mistral Small 3.1 | Narrow decision boundary |
|---|---|---|---|
| Routing · 200 / 30 | $3.45 | $4.80 | 10% accepted-result uplift required |
| Extraction · 1,000 / 250 | $18.75 | $30.00 | 15% accepted-result uplift required |
Formula: requests × (input tokens × input $/M + output tokens × output $/M) ÷ 1,000,000. The uplift is a planning threshold, not a measured quality claim.
3. API price/latency versus edge-deployment inputs
| Option | Dated calculation | Input ledger | Boundary |
|---|---|---|---|
| Ministral API | $18.75 | 158 tokens/sec; TTFT 210 ms; 5 measured samples | API cost is sourced |
| Edge deployment | Unavailable | Hardware, utilization, energy, image units, cache, batch | Collect user inputs before comparing |
Verified 2026-08-14. “Unavailable” means no compatible dated evidence was found; it is never treated as zero. First-party price source · Run this scenario.
All three decisions below are computed for Ministral 8B; fixed inputs, formulas, dated sources, speed sample state, and unavailable mechanics are visible.
Exact-model pricing guide · verified 2026-09-08
Exact model boundary: Mistral ministral-8b (slug ministral-8b). First-party provider pricing and API documentation remain fact owners.
Symmetric token pricing and high-volume spend matrix
Frozen scenario board. Formula / deterministic rule: monthly_spend = calls * ((in_tokens * 0.15 + out_tokens * 0.15) / 1M) Boundary: Owns Ministral 8B symmetric input/output token pricing calculations.
| Frozen scenario | Model, identity, provider, and evidence fields | Result | State |
|---|---|---|---|
| 100K compact classification queries | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=100K compact classification queries; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 100K compact classification queries is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 250K agent reasoning step checks | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=250K agent reasoning step checks; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 250K agent reasoning step checks is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 1M high-volume IoT telemetry passes | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=1M high-volume IoT telemetry passes; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 1M high-volume IoT telemetry passes is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| long-output code generation pass | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=long-output code generation pass; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — long-output code generation pass is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| batch processing queue (50% discount) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=batch processing queue (50% discount); workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — batch processing queue (50% discount) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unresolved billing currency | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unresolved billing currency; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unresolved billing currency has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Mistral AI model pricing; verification date 2026-09-08. Missing or conflicting joins fail closed.
Edge on-device versus managed cloud API latency audit
Frozen scenario board. Formula / deterministic rule: turnaround = ttft + (tokens_out / tps); edge provides zero network overhead Boundary: Owns on-device quantization and cloud API response time benchmarks.
| Frozen scenario | Model, identity, provider, and evidence fields | Result | State |
|---|---|---|---|
| sub-50ms local edge classification | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=sub-50ms local edge classification; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — sub-50ms local edge classification is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| smart appliance embedded assistant (<100ms) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=smart appliance embedded assistant (<100ms); environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — smart appliance embedded assistant (<100ms) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| cloud API call with network transit (<250ms) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=cloud API call with network transit (<250ms); environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — cloud API call with network transit (<250ms) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| high-concurrency edge sensor stream | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=high-concurrency edge sensor stream; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — high-concurrency edge sensor stream is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| quantization accuracy degradation canary | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=quantization accuracy degradation canary; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — quantization accuracy degradation canary is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unsupported embedded runtime | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unsupported embedded runtime; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unsupported embedded runtime has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Mistral AI documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
Local VRAM deployment break-even: RTX 4090 vs Cloud API
Frozen scenario board. Formula / deterministic rule: cloud_spend = monthly_tokens * 0.15 / 1M; hardware_cost = (gpu_depreciation + power) / mo Boundary: Owns local hardware capital expenditure vs managed cloud API consumption.
| Frozen scenario | Model, identity, provider, and evidence fields | Result | State |
|---|---|---|---|
| low edge volume (<5M tokens/mo) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=low edge volume (<5M tokens/mo); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — low edge volume (<5M tokens/mo) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 25M monthly tokens (cloud optimal) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=25M monthly tokens (cloud optimal); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 25M monthly tokens (cloud optimal) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 100M tokens/month (hardware break-even) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=100M tokens/month (hardware break-even); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 100M tokens/month (hardware break-even) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 500M high-volume saturation (local optimal) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=500M high-volume saturation (local optimal); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 500M high-volume saturation (local optimal) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unplanned hardware failure and downtime | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unplanned hardware failure and downtime; monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unplanned hardware failure and downtime has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
| unresolved local electricity tariff | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unresolved local electricity tariff; monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unresolved local electricity tariff has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Mistral AI model pricing; verification date 2026-09-08. Missing or conflicting joins fail closed.
Method and limitations: formulas are deterministic; observed and assumed inputs are labeled; no missing provider, host, account, region, realm, alias, snapshot, revision, weight, artifact, control, tool, modality, workload, rate-period, timestamp, or result is transferred. Run this scenario →
Verified Model Pricing Intelligence•Audit date: 2026-09-08
Ministral 8B API Pricing: Symmetric Token Rates for Edge & Cloud
Ministral 8B costs $0.15 per million input tokens and $0.15 per million output tokens ($0.15/M blended at 3:1). Features unique symmetric pricing, 128K context, and ultra-compact edge deployment. Verified 2026-09-08.
Symmetric token pricing and high-volume spend matrix
Frozen scenario board. Formula / deterministic rule: monthly_spend = calls * ((in_tokens * 0.15 + out_tokens * 0.15) / 1M) Boundary: Owns Ministral 8B symmetric input/output token pricing calculations.
| Frozen scenario | Model, identity, provider, and evidence fields | Result | State |
|---|---|---|---|
| 100K compact classification queries | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=100K compact classification queries; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 100K compact classification queries is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 250K agent reasoning step checks | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=250K agent reasoning step checks; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 250K agent reasoning step checks is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 1M high-volume IoT telemetry passes | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=1M high-volume IoT telemetry passes; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 1M high-volume IoT telemetry passes is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| long-output code generation pass | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=long-output code generation pass; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — long-output code generation pass is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| batch processing queue (50% discount) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=batch processing queue (50% discount); workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — batch processing queue (50% discount) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unresolved billing currency | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unresolved billing currency; workload; prompt tokens; completion tokens; input cost; output cost; monthly spend; efficiency verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unresolved billing currency has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Mistral AI model pricing; verification date 2026-09-08. Missing or conflicting joins fail closed.
Edge on-device versus managed cloud API latency audit
Frozen scenario board. Formula / deterministic rule: turnaround = ttft + (tokens_out / tps); edge provides zero network overhead Boundary: Owns on-device quantization and cloud API response time benchmarks.
| Frozen scenario | Model, identity, provider, and evidence fields | Result | State |
|---|---|---|---|
| sub-50ms local edge classification | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=sub-50ms local edge classification; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — sub-50ms local edge classification is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| smart appliance embedded assistant (<100ms) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=smart appliance embedded assistant (<100ms); environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — smart appliance embedded assistant (<100ms) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| cloud API call with network transit (<250ms) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=cloud API call with network transit (<250ms); environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — cloud API call with network transit (<250ms) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| high-concurrency edge sensor stream | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=high-concurrency edge sensor stream; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — high-concurrency edge sensor stream is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| quantization accuracy degradation canary | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=quantization accuracy degradation canary; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — quantization accuracy degradation canary is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unsupported embedded runtime | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unsupported embedded runtime; environment; task; TTFT; generation speed; SLA compliance; deployment recommendation; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unsupported embedded runtime has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Mistral AI documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
Local VRAM deployment break-even: RTX 4090 vs Cloud API
Frozen scenario board. Formula / deterministic rule: cloud_spend = monthly_tokens * 0.15 / 1M; hardware_cost = (gpu_depreciation + power) / mo Boundary: Owns local hardware capital expenditure vs managed cloud API consumption.
| Frozen scenario | Model, identity, provider, and evidence fields | Result | State |
|---|---|---|---|
| low edge volume (<5M tokens/mo) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=low edge volume (<5M tokens/mo); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — low edge volume (<5M tokens/mo) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 25M monthly tokens (cloud optimal) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=25M monthly tokens (cloud optimal); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 25M monthly tokens (cloud optimal) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 100M tokens/month (hardware break-even) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=100M tokens/month (hardware break-even); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 100M tokens/month (hardware break-even) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 500M high-volume saturation (local optimal) | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=500M high-volume saturation (local optimal); monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — 500M high-volume saturation (local optimal) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unplanned hardware failure and downtime | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unplanned hardware failure and downtime; monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unplanned hardware failure and downtime has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
| unresolved local electricity tariff | model=ministral-8b; slug=ministral-8b; provider=Mistral; scenario=unresolved local electricity tariff; monthly token volume; cloud API spend; local RTX 4090 monthly TCO; financial break-even; deployment verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-08; measurement versus assumption=explicit | Unavailable — unresolved local electricity tariff has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Mistral AI model pricing; verification date 2026-09-08. Missing or conflicting joins fail closed.
Exact-Model Pricing Evidence•Verified 2026-08-14; revalidation required before current claims.
Ministral 8B pricing evidence
Source-backed dated rate shape: dated input=$0.150000/M; output=$0.150000/M. Ministral 8B exact-model unit economics only; Mistral residency, local hardware guarantees, and cheapest rankings retain their owners. Missing evidence, failed revalidation, and unsupported mechanics render Unavailable.
Module 1 of 3: Tagging and routing volume ladder
Novel contribution boundary: Owns dated Ministral 8B arithmetic for fixed high-volume token shapes; edge latency and local deployment are not asserted. Formula / deterministic rule: spend = requests × (inputTokens × inputCostPer1k + outputTokens × outputCostPer1k) / 1,000
| Scenario | Exact model, provider, dated rates, and fixed inputs | Result | State |
|---|---|---|---|
| 100K tagging requests | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=requests=100,000; inputTokens=300; outputTokens=80; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| 1M routing requests | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=requests=1,000,000; inputTokens=600; outputTokens=120; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| 10M compact extraction requests | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=requests=10,000,000; inputTokens=1,000; outputTokens=200; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
First-party provenance: https://docs.mistral.ai/models/model-cards/ministral-3-8b-25-12; registry provider mistral; verified 2026-08-14; freshness gate=2026-09-14. Revalidate before any current-price claim.
Module 2 of 3: Ministral→Small output-cost boundary
Novel contribution boundary: Owns dated cross-model cost arithmetic with explicit retry input; accepted-result quality and edge/cloud routing are not sourced. Formula / deterministic rule: requiredAcceptedRate = SmallAttemptCost / MinistralAttemptCost; observed acceptance = Unavailable
| Scenario | Exact model, provider, dated rates, and fixed inputs | Result | State |
|---|---|---|---|
| 300-token tag; 10% retry share; 100 retry-input tokens | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=requests=1; inputTokens=300; outputTokens=80; retryInputTokens=100; retryShare=10%; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| 600-token route; 20% retry share; 250 retry-input tokens | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=requests=1; inputTokens=600; outputTokens=120; retryInputTokens=250; retryShare=20%; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| 1K-token extraction; 30% retry share; 500 retry-input tokens | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=requests=1; inputTokens=1,000; outputTokens=200; retryInputTokens=500; retryShare=30%; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
First-party provenance: https://docs.mistral.ai/models/model-cards/ministral-3-8b-25-12; registry provider mistral; verified 2026-08-14; freshness gate=2026-09-14. Revalidate before any current-price claim.
Module 3 of 3: API-versus-edge evidence ledger
Novel contribution boundary: Owns exact-model Ministral registry evidence only; hardware, energy, cache, batch, and image behavior are not fabricated. Formula / deterministic rule: evidence = exact-model registry field when present; hardware, energy, cache, batch, and image inputs = Unavailable
| Scenario | Exact model, provider, dated rates, and fixed inputs | Result | State |
|---|---|---|---|
| Local hardware or VRAM requirement | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=evidenceField=hardware; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| Cache or batch treatment | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=evidenceField=cache; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| Image input unit or accounting | model=ministral-8b; provider=mistral; registryRates=dated input=$0.150000/M; output=$0.150000/M; comparisonModel=mistral-small; fixedInputs=evidenceField=image-unit; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-08-14, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
First-party provenance: https://docs.mistral.ai/models/model-cards/ministral-3-8b-25-12; registry provider mistral; verified 2026-08-14; freshness gate=2026-09-14. Revalidate before any current-price claim.
How fast is Ministral 8B?
Tokens / sec
158
TTFT
210 ms
Rank
#8 of 31
$ / M ÷ t/s
$0.0009
Measured with 5 runs on a fixed prompt — see the full methodology.
How much does Ministral 8B cost at scale?
| Tokens / month | Est. cost (blended 3:1) |
|---|---|
| 100,000 | $0.01 |
| 1,000,000 | $0.15 |
| 10,000,000 | $1.50 |
| 100,000,000 | $15.00 |
How does Ministral 8B compare with other models?
Mistral Small 3.1 — $0.26/MCodestral — $0.45/MMistral Large 3 — $0.75/MMistral Medium 3 — $3.00/MGPT-5 Nano — $0.14/MGPT-OSS 20B — $0.13/MMuse Spark 1.3 Contributor — $0.13/M
What is Ministral 8B best for?
What should you explore next for Ministral 8B?
What are common questions about Ministral 8B?
Is Ministral 8B cheaper than GPT-5 Nano?
Ministral 8B costs $0.15/M blended tokens, GPT-5 Nano costs $0.14/M — GPT-5 Nano is cheaper.
How much does 1 million tokens cost with Ministral 8B?
At a 3:1 input:output ratio, 1 million blended tokens costs approximately $0.15. Pure input costs $0.15/M; pure output costs $0.15/M.
What does Ministral 8B cost at high volume?
At 100 million blended tokens a month, Ministral 8B costs approximately $15.00. See the cost-at-scale table below for other volumes.