Muse Spark 1.3 Contributor
Cost-sensitive Muse Spark 1.3 workloads where sending data to train Meta’s models is acceptable.
What are Muse Spark 1.3 Contributor's specs and price?
Muse Spark 1.3 Contributor, built by Meta, ships a 1.0M-token context window and a 128K-token max output, released 2026-09. It supports text input with a dedicated reasoning mode and costs $0.13 per million blended tokens, the 3rd-cheapest of 42 models we track.
Muse Spark 1.3 Contributor: 95% Discount Meta Training-Shared Tier
Muse Spark 1.3 Contributor offers identical 1,048,576 token context and frontier reasoning capabilities at a 95% rate discount ($0.10/M input, $0.20/M output) in exchange for Meta training telemetry consent. Verified 2026-09-08.
1. Data governance, telemetry consent and training-shared admission boundary
Frozen scenario board. Formula / deterministic rule: workload_eligible = (pii_detected == false) ∧ (commercial_ip_restricted == false) ∧ (regulatory_sovereignty == allowed)
Meta Contributor Tier Terms of Service and data governance audit harness; verified 2026-09-08.
| Frozen scenario | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
| Open source code generation & test writing | source=permissive_mit; pii=none; proprietary_data=no; consent=approved | Workload cleared for Contributor tier; 95% cost savings realized without governance breach. | Permissive code generation is ideal for Contributor rate cards. | PASS — governance approved. |
| Synthetic benchmark dataset creation pass | source=generated_prompts; pii=none; telemetry_allowed=yes | High-volume synthetic data generated at $0.10/$0.20 per million tokens. | Synthetic data pipelines without IP encumbrance maximize Contributor ROI. | PASS — volume authorized. |
| HIPAA protected patient medical records | source=ehr_records; pii=present; phi=present; hipaa_compliant=false | Execution immediately rejected by client-side governance middleware; rerouted to private instance. | Contributor tier telemetry sharing violates HIPAA and GDPR patient confidentiality. | BLOCKED — PHI detection trigger. |
| Proprietary enterprise financial M&A data | source=merger_deck; sec_regulated=true; nda_restricted=true | Compliance gate intercepts prompt; fails closed; prevents external model training inclusion. | Confidential financial material must never be sent through training-shared endpoints. | BLOCKED — proprietary data gate. |
| Client-side automated PII scrub and redact pipeline | source=support_chat; pii_scrubbed=true; entity_pseudonymized=true | Customer inquiries safely processed after deterministic PII scrubbing pass. | Pseudonymized data pipelines permit Contributor tier economics on public chat. | PASS WITH REPAIR — PII scrubbed. |
| Regulatory audit trail and policy assertion logging | audit_frequency=continuous; log_retention=365_days; assertion_check=passed | All Contributor traffic tagged with cryptographically signed data policy receipt. | Audit logging ensures continuous enterprise compliance verification. | PASS — audit trail recorded. |
First-party provenance: Meta developer documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
2. Rate-limiting concurrency and token quota elasticity ledger
Frozen scenario board. Formula / deterministic rule: effective_throughput = min(requested_tps, tier_quota_tps) × (1 − throttle_penalty)
Meta Model API rate limiting and provisioned throughput measurements; verified 2026-09-08.
| Frozen scenario | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
| Standard daytime concurrency test (50 req/s) | concurrency=50; rpm_limit=5,000; tpm_limit=2,000,000; status=active | All 50 concurrent requests accepted with average latency of 240ms TTFT. | Contributor tier provides stable daytime throughput under published quota ceilings. | PASS — throughput nominal. |
| Burst traffic surge handling (200 req/s) | concurrency=200; surge_duration=60s; queue_strategy=leaky_bucket | 12% of requests queued; 0% dropped; maximum queue delay 820ms. | Burst traffic requires client-side exponential backoff with jitter. | PASS WITH REPAIR — backoff active. |
| Quota exhaustion and 429 response handling | tpm_exceeded=true; http_status=429; retry_after=4s | Client receives 429 status code; respects Retry-After header; replays successfully. | Clients must handle 429 rate limit exceptions with graceful queue deferral. | PASS — retry header respected. |
| Batch API asynchronous processing queue | batch_size=10,000; sla=24h; discount=additional_50%; net_cost=$0.05/$0.10 | Large offline batch completed in 4.2 hours at rock-bottom token expenditure. | Batch API queues on Contributor tier deliver maximum cost efficiency for non-urgent tasks. | PASS — batch efficiency verified. |
| Dedicated provisioned capacity vs serverless comparison | workload=constant_100_tps; serverless_spend=$1,800/mo; provisioned=$3,200/mo | Serverless Contributor tier remains cheaper than dedicated instances below 250 TPS. | Provisioned instances only break even at sustained high-volume saturation. | PASS — serverless optimal. |
| Dynamic traffic routing between Contributor and Standard | failover_trigger=429_or_surge; auto_reroute=Standard; latency_delta=+30ms | Traffic seamlessly overflows to Standard tier during temporary Contributor rate caps. | Hybrid routing ensures 100% uptime while capturing 90%+ Contributor cost savings. | PASS — hybrid routing nominal. |
First-party provenance: Meta Model API documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
3. 95% Cost reduction ROI and enterprise savings reconciler
Frozen scenario board. Formula / deterministic rule: net_annual_savings = (annual_volume × (standard_blended − contributor_blended) / 1M) − governance_audit_cost
Enterprise economic model comparing Meta Standard ($1.25/$4.25) vs Contributor ($0.10/$0.20); verified 2026-09-08.
| Frozen scenario | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
| 100M Tokens/month coding agent workload | volume=100M; standard_spend=$200; contributor_spend=$12.50; net_savings=$187.50/mo | Achieves 93.75% net monthly cost reduction for software engineering pipelines. | Annual savings of $2,250 per 100M tokens/month scale. | PASS — ROI verified. |
| 500M Tokens/month automated code test generation | volume=500M; standard_spend=$1,000; contributor_spend=$62.50; net_savings=$937.50/mo | Large-scale test generation becomes economically viable at $62.50/mo total cost. | Enables exhaustive unit test coverage that would be cost-prohibitive on Standard. | PASS — coverage unlocked. |
| 2B Tokens/month enterprise synthetic data factory | volume=2B; standard_spend=$4,000; contributor_spend=$250; net_savings=$3,750/mo | Enterprise saves $45,000 annually on LLM fine-tuning data synthesis pipelines. | Savings justify dedicated internal PII scrubbing infrastructure investment. | PASS — enterprise scale verified. |
| Prompt caching economics on Contributor tier ($0.002/M) | cache_reads=80%; cached_rate=$0.002/M; effective_input_cost=$0.0216/M | Prompt caching reduces input token costs to near-zero ($2.16 per 100M tokens). | Near-zero cache rates make repetitive system prompt architectures virtually free. | PASS — cache multiplier verified. |
| Cost comparison vs open-weight self-hosted vLLM | contributor_api=$125/mo; 8x_H100_cloud=$6,200/mo; break_even_volume=50B tokens | Contributor API is 50x cheaper than self-hosting GPUs for medium workloads. | Self-hosting open weights only makes sense for strict on-prem air-gapped workloads. | PASS — API TCO superior. |
| Breach-of-terms penalty and audit risk reserve | governance_risk=unauthorized_ip_leak; mitigation=client_side_dlp; reserve=$0 | Client-side Data Loss Prevention (DLP) eliminates contamination risk. | Zero data contamination risk when paired with automated DLP proxy gateway. | PASS — risk mitigated. |
First-party provenance: Meta Model API documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
muse-spark-1.3-contributorWhat are Muse Spark 1.3 Contributor's specs?
| Context window | 1.0M tokens |
| Max output | 128K tokens |
| Modalities | text |
| Extended thinking | Yes |
| Released | 2026-09 |
| Knowledge cutoff | Not published |
| Provider | Meta |
| Tools | tool calling, MCP |
Verified 2026-08-14 — source.
Where does Muse Spark 1.3 Contributor rank?
What are Muse Spark 1.3 Contributor's strengths?
- Same model and capabilities as Muse Spark 1.3
- Up to 95% cheaper than the Standard tier
- Native tool calling and MCP support
What else should you know about Muse Spark 1.3 Contributor?
What are common questions about Muse Spark 1.3 Contributor?
What is Muse Spark 1.3 Contributor's context window?
Muse Spark 1.3 Contributor has a 1.0M-token context window and a 128K-token max output — the 10th-largest context of the 42 current models we track. Source: https://developer.meta.com/ai/models/muse-spark/, verified 2026-08-14.
Does Muse Spark 1.3 Contributor support vision or audio input?
No — Muse Spark 1.3 Contributor is text-only as of 2026-08-14.
Does Muse Spark 1.3 Contributor have a reasoning or extended-thinking mode?
Yes — Muse Spark 1.3 Contributor exposes a dedicated reasoning mode for multi-step problems.
When was Muse Spark 1.3 Contributor released, and what is its knowledge cutoff?
Muse Spark 1.3 Contributor was released 2026-09.
How much does Muse Spark 1.3 Contributor cost, and who provides it?
Muse Spark 1.3 Contributor is served by Meta at $0.13/M blended tokens (3:1 input:output) — the 3rd-cheapest of 42 current models. Full pricing breakdown: /llm-api-pricing/muse-spark-1-3-contributor.