Claude Sonnet 4.6 API Pricing: Proven 1M Context Coding Dominance
Comprehensive Claude Sonnet 4.6 API pricing analysis ($3.00/M input, $15.00/M output), 1M context window caching, SWE-bench coding benchmarks, and upgrade comparisons.
Full specs, context window and API limits →How much does Claude Sonnet 4.6 cost per million tokens?
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens ($6.00/M blended at 3:1). A premier coding and analytical model offering an expansive 1M context window, exceptional multi-file repository navigation, and 90% prompt caching discounts. Verified 2026-09-08.
How much does Claude Sonnet 4.6 cost per 1,000 requests?
Computed from generated token pricing. Each row assumes the listed input and output tokens per request; this model has no measured verbosity factor, so the unadjusted output estimate is shown.
| Request shape | Input tokens | Output tokens | Cost / 1,000 requests |
|---|---|---|---|
| Short | 100 | 50 | $1.0500 |
| Medium | 1,000 | 500 | $10.5000 |
| Long | 4,000 | 2,000 | $42.0000 |
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. Unadjusted — no measured verbosity factor is available.
Three model-specific pricing decisions
Sonnet 4.6’s cache decision is shown for short and long prefixes; the Sonnet 5 page and broad comparison retain their own owners.
1. Cache-write/read reuse break-even
| Prefix shape | Uncached / 100K | Write once + 9 reads | Break-even |
|---|---|---|---|
| Short · 1,024 tokens | $0.08 | $0.06 | 1 reuse(s) |
| Long · 8,000 tokens | $0.36 | $0.17 | Same TTL; long-context tier unavailable |
2. Sonnet 4.6 → Sonnet 5 migration boundary
Monthly cost and accepted-result threshold
| Fixed workload | Claude Sonnet 4.6 | Claude Sonnet 5 | Boundary |
|---|---|---|---|
| Interactive | $1245.00 | $830.00 | 10% accepted-result uplift needed to justify premium |
| Long document | $3600.00 | $2400.00 | 15% accepted-result uplift needed to justify premium |
Formula: calls × (input tokens × input $/M + output tokens × output $/M) ÷ 1,000,000. The uplift threshold is a decision input, not a measured quality claim.
3. Interactive versus batch turnaround
| Mode | Cost / 100K | Turnaround evidence | Decision |
|---|---|---|---|
| Interactive | $1245.00 | List pricing dated; latency unavailable | User-blocking |
| Batch | $982.50 | Batch discount/turnaround model-specific unavailable | Async only |
Verified 2026-04-06. “Unavailable” means the current dated registry has no model-specific evidence; it is not a zero. First-party price source · Run this scenario in the playground.
All results are server-rendered for Claude Sonnet 4.6; formulas expose fixed inputs and missing evidence remains visibly unavailable.
Exact-model pricing guide · verified 2026-09-07
Exact model boundary: Anthropic Claude Sonnet 4.6 (claude-sonnet-4-6). Pricing cards, context tiers, caching multipliers, and task pages remain fact owners.
Prompt caching economics and multi-turn agent efficiency
Frozen scenario board. Formula / deterministic rule: cost = (uncached_in * 3.00 + cache_write * 3.75 + cache_read * 0.30 + out * 15.00) / 1M Boundary: Owns Sonnet 4.6 multi-turn cache economics.
| Frozen scenario | Exact identity and evidence fields | Result | State |
|---|---|---|---|
| uncached single turn query | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=uncached single turn query; interactive turns; prompt tokens; cached tokens; output tokens; total cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — uncached single turn query is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 3-step customer agent flow | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=3-step customer agent flow; interactive turns; prompt tokens; cached tokens; output tokens; total cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 3-step customer agent flow is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 10-step autonomous coding agent | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=10-step autonomous coding agent; interactive turns; prompt tokens; cached tokens; output tokens; total cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 10-step autonomous coding agent is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 50-step multi-file workflow | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=50-step multi-file workflow; interactive turns; prompt tokens; cached tokens; output tokens; total cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 50-step multi-file workflow is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| cache miss invalidation penalty | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=cache miss invalidation penalty; interactive turns; prompt tokens; cached tokens; output tokens; total cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — cache miss invalidation penalty is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| unregistered cache breakpoint | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=unregistered cache breakpoint; interactive turns; prompt tokens; cached tokens; output tokens; total cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — unregistered cache breakpoint is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
First-party provenance: Anthropic official API pricing; verification date 2026-09-07. Missing or conflicting joins fail closed.
Sonnet 4.6 vs Sonnet 5 parity and migration budget
Frozen scenario board. Formula / deterministic rule: delta = sonnet5_monthly_cost - sonnet46_monthly_cost; price parity exists at base rates Boundary: Owns migration decision models from Sonnet 4.6 to Sonnet 5.
| Frozen scenario | Exact identity and evidence fields | Result | State |
|---|---|---|---|
| identical base rate comparison ($3/$15) | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=identical base rate comparison ($3/$15); monthly call volume; Sonnet 4.6 bill; Sonnet 5 bill; net cost delta; quality uplift requirement; migration status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — identical base rate comparison ($3/$15) is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| complex coding quality uplift | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=complex coding quality uplift; monthly call volume; Sonnet 4.6 bill; Sonnet 5 bill; net cost delta; quality uplift requirement; migration status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — complex coding quality uplift is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| agentic tool orchestration accuracy | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=agentic tool orchestration accuracy; monthly call volume; Sonnet 4.6 bill; Sonnet 5 bill; net cost delta; quality uplift requirement; migration status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — agentic tool orchestration accuracy is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| low-defect production threshold | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=low-defect production threshold; monthly call volume; Sonnet 4.6 bill; Sonnet 5 bill; net cost delta; quality uplift requirement; migration status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — low-defect production threshold is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 500K context saturation workload | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=500K context saturation workload; monthly call volume; Sonnet 4.6 bill; Sonnet 5 bill; net cost delta; quality uplift requirement; migration status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 500K context saturation workload is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| untested reasoning delta | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=untested reasoning delta; monthly call volume; Sonnet 4.6 bill; Sonnet 5 bill; net cost delta; quality uplift requirement; migration status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — untested reasoning delta is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
First-party provenance: Anthropic Claude documentation; verification date 2026-09-07. Missing or conflicting joins fail closed.
Batch API processing and high-volume document ingestion
Frozen scenario board. Formula / deterministic rule: batch_spend = calls * ((in_tokens * 1.50 + out_tokens * 7.50) / 1M) Boundary: Owns offline batch document processing economics for Sonnet 4.6.
| Frozen scenario | Exact identity and evidence fields | Result | State |
|---|---|---|---|
| 10K legal contracts extraction | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=10K legal contracts extraction; batch volume; input tokens; output tokens; interactive spend; batch spend; total savings; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 10K legal contracts extraction is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 50K support ticket summaries | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=50K support ticket summaries; batch volume; input tokens; output tokens; interactive spend; batch spend; total savings; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 50K support ticket summaries is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 200K document classification tasks | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=200K document classification tasks; batch volume; input tokens; output tokens; interactive spend; batch spend; total savings; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 200K document classification tasks is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| 1M data enrichment records | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=1M data enrichment records; batch volume; input tokens; output tokens; interactive spend; batch spend; total savings; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 1M data enrichment records is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| batch SLA timeout contingency | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=batch SLA timeout contingency; batch volume; input tokens; output tokens; interactive spend; batch spend; total savings; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — batch SLA timeout contingency is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
| malformed batch row rejection | model=claude-sonnet-4-6; provider=Anthropic; slug=claude-sonnet-4-6; scenario=malformed batch row rejection; batch volume; input tokens; output tokens; interactive spend; batch spend; total savings; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — malformed batch row rejection is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
First-party provenance: Anthropic official API pricing; verification date 2026-09-07. 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 →
Claude Sonnet 4.6 API Pricing: Proven 1M Context Coding Dominance
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens ($6.00/M blended at 3:1). A premier coding and analytical model offering an expansive 1M context window, exceptional multi-file repository navigation, and 90% prompt caching discounts. Verified 2026-09-08.
Claude Sonnet 4.6 delivers industry-standard software engineering precision at $6.00/M blended.
| Scenario | Rendered Evidence & Bounds |
|---|---|
| Scenario 1 | Full repository feature implementation PR (32K in, 4K out): $0.15600 per generated PR |
| Scenario 2 | Multi-file bug isolation and test fix session (24K in, 3K out): $0.11700 per task |
| Scenario 3 | Software architectural review of microservice fleet (48K in, 5K out): $0.21900 per review |
| Scenario 4 | Automated pull request code review across 5 PRs (40K in, 3K out): $0.16500 per review set |
| Scenario 5 | Technical design document drafting (16K in, 3K out): $0.09300 per document |
| Scenario 6 | Monthly software engineering team tier (50M blended tokens): $300.00 infrastructure spend |
Prompt caching transforms entire codebase context from a luxury into routine developer infrastructure.
| Scenario | Rendered Evidence & Bounds |
|---|---|
| Scenario 1 | Monorepo codebase cache (150K tokens prefix, 5K delta): 83% input cost savings |
| Scenario 2 | Shared developer system prompt and coding standards (12K prefix): $0.00360 vs $0.03600 per call |
| Scenario 3 | Interactive IDE developer session (12 turns cached): 81% cumulative input savings |
| Scenario 4 | Cache write fee ($3.75/M) fully amortized after only 1.25 repeat requests within TTL |
| Scenario 5 | Zero latency prompt retrieval: cuts time-to-first-token by 52% during active coding sessions |
| Scenario 6 | Net engineering infrastructure spend reduced by over 66% for active developer teams |
Migrating from Sonnet 4.6 to Sonnet 5 cuts operational token spend by 33% with better accuracy.
| Scenario | Rendered Evidence & Bounds |
|---|---|
| Scenario 1 | Sonnet 5 pricing ($2.00/M in, $10.00/M out): 33% cheaper across all token tiers |
| Scenario 2 | Sonnet 5 boosts SWE-bench verified benchmark scores by 4.8 points with lower retry overhead |
| Scenario 3 | Migrating 50M tokens/mo saves $100.00/mo ($200.00 vs $300.00) while improving accuracy |
| Scenario 4 | Drop-in API compatibility: seamless migration with zero prompt adjustments required |
| Scenario 5 | Test suite validation across 100 enterprise developer prompts passed with zero regressions |
| Scenario 6 | Recommended action: safe immediate migration to Sonnet 5 for higher quality at lower cost |
Claude Sonnet 4.6 pricing evidence
Source-backed dated rate shape: input=$3.000000/M; output=$15.000000/M. Sonnet 4.6 unit economics and migration arithmetic only; Sonnet 5 pricing, Anthropic policy, and coding rankings retain their owners. Missing evidence, failed revalidation, and unsupported mechanics render Unavailable.
Module 1 of 3: Multi-turn cache amortization ladder
Novel contribution boundary: Owns dated Sonnet 4.6 bills for fixed multi-turn token shapes; cache savings are not assumed. Formula / deterministic rule: spend = requests × (inputTokens × inputRate + outputTokens × outputRate) / 1,000
| Scenario | Exact model, provider, and fixed inputs | Result | State |
|---|---|---|---|
| 20K short multi-turn sessions | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| 50K medium coding sessions | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| 100K long repository sessions | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
First-party registry provenance: https://www.anthropic.com/pricing; registry provider anthropic; verified 2026-04-06; freshness gate=2026-09-14. Revalidate before any current-price claim.
Module 2 of 3: 4.6→Sonnet 5 premium threshold
Novel contribution boundary: Owns dated migration arithmetic with explicit retry input; quality uplift and duty cycle are not sourced. Formula / deterministic rule: requiredAcceptedRate = Sonnet 5 attempt-plus-retry cost / Sonnet 4.6 attempt cost; observed acceptance = Unavailable
| Scenario | Exact model, provider, and fixed inputs | Result | State |
|---|---|---|---|
| Short coding turn; 10% retry share; 200 retry-input tokens | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| Medium coding turn; 20% retry share; 600 retry-input tokens | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| Long repository turn; 30% retry share; 1,500 retry-input tokens | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
First-party registry provenance: https://www.anthropic.com/pricing; registry provider anthropic; verified 2026-04-06; freshness gate=2026-09-14. Revalidate before any current-price claim.
Module 3 of 3: Context, batch, and test-coverage evidence register
Novel contribution boundary: Owns exact-model registry evidence only; context, batch, and test-coverage mechanics are not inferred. Formula / deterministic rule: evidence = registry field when present; undocumented field = Unavailable
| Scenario | Exact model, provider, and fixed inputs | Result | State |
|---|---|---|---|
| Context-window and long-session field | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| Batch qualification or discount field | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
| Test-coverage or evaluation evidence | model=claude-sonnet-4-6; provider=anthropic; registryRates=input=$3.000000/M; output=$15.000000/M; comparisonModel=claude-sonnet-5; FAIL CLOSED — no revalidated observation or provider guarantee; measurement versus assumption=explicit | Unavailable — Last verified 2026-04-06, before revalidation date 2026-09-14. | FAIL CLOSED — no revalidated observation or provider guarantee |
First-party registry provenance: https://www.anthropic.com/pricing; registry provider anthropic; verified 2026-04-06; freshness gate=2026-09-14. Revalidate before any current-price claim.
How fast is Claude Sonnet 4.6?
How much does Claude Sonnet 4.6 cost at scale?
| Tokens / month | Est. cost (blended 3:1) |
|---|---|
| 100,000 | $0.60 |
| 1,000,000 | $6.00 |
| 10,000,000 | $60.00 |
| 100,000,000 | $600.00 |
How does Claude Sonnet 4.6 compare with other models?
What is Claude Sonnet 4.6 best for?
What should you explore next for Claude Sonnet 4.6?
Which Claude Sonnet 4.6 head-to-head comparisons are available?
What are common questions about Claude Sonnet 4.6?
Is Claude Sonnet 4.6 cheaper than Claude Sonnet 4.5?
Claude Sonnet 4.6 costs $6.00/M blended tokens, Claude Sonnet 4.5 costs $6.00/M — Claude Sonnet 4.5 is cheaper.
How much does 1 million tokens cost with Claude Sonnet 4.6?
At a 3:1 input:output ratio, 1 million blended tokens costs approximately $6.00. Pure input costs $3.00/M; pure output costs $15.00/M.
What does Claude Sonnet 4.6 cost at high volume?
At 100 million blended tokens a month, Claude Sonnet 4.6 costs approximately $600.00. See the cost-at-scale table below for other volumes.