# Muse Spark 1.2

Meta, released 2026-08-05

**Verdict: conditional.** Meta's coding-focused model update, co-trained with the Muse Code harness — 82.9% on Terminal-Bench 2.1 and 59.3% on DeepSWE 1.1, second only to Claude Opus 5 on the former and trailing both Opus 5 and GPT-5.6 Terra on the latter. Standard pricing unchanged from 1.1 at $1.25 input / $4.25 output per 1M tokens; contributor tier drops to $0.10/$0.20 in exchange for training-data rights. All published benchmarks remain vendor-run with no independent reproduction. Best for cost-sensitive coding teams willing to test a model explicitly optimized for its own harness. Last verified 2026-10-13.

## Pricing

- Input: $1.25 / 1M tokens
- Output: $4.25 / 1M tokens
- Context: 1M tokens

## Strengths

- Cost-sensitive coding agent workloads: $1.25/$4.25 standard pricing undercuts Claude, GPT, and Gemini flagships by 3-5x
- Muse Code agent integration: co-trained with the harness for best-in-tool performance
- Long-horizon autonomous coding: 1,000+ tool calls over 24 hours, event-log crash recovery, sustained optimization past initial exploration
- Multimodal repo understanding: accepts text, image, video, and PDF inputs for coding context
- Production coding accuracy without independent validation: all published benchmarks are vendor-run in Meta's own framework; no third-party reproduction exists yet
- Non-coding or general agentic tasks: coding-focused checkpoint; reasoning, knowledge, math, and multilingual benchmarks remain unpublished
- Open-weight or self-hosting requirements: closed weights, API-only; no Hugging Face weights, no fine-tuning
- High-concurrency production on contributor tier: 100 RPM cap vs 3,000 standard per Meta's rate-limit docs; training-data grant on all contributor traffic

## Benchmarks

- Terminal-Bench 2.1: 82.9% (https://research.meta.ai/static/muse-spark-1-2-methodology)
- DeepSWE 1.1: 59.3% (https://research.meta.ai/static/muse-spark-1-2-methodology)
- Meta Internal Coding Bench: 70.6% (https://research.meta.ai/static/muse-spark-1-2-methodology)
- BenchLM Aggregate: 60.3/100 (#49 of 216) (https://benchlm.ai/models/muse-spark-1-2)

## Sources

- [Meta Muse Spark 1.2 evaluation methodology (official PDF)](https://research.meta.ai/static/muse-spark-1-2-methodology)
- [Meta AI Research Blog — Muse Code and Muse Spark 1.2 announcement (official)](https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2)
- [Meta Model API docs: models (Muse Spark Open listed, 816B MoE with 42B active)](https://dev.meta.ai/docs/models)
- [BenchLM.ai — Muse Spark 1.2 profile](https://benchlm.ai/models/muse-spark-1-2)
- [VentureBeat: Meta enters AI coding wars](https://venturebeat.com/orchestration/meta-enters-the-ai-coding-wars-with-muse-spark-1-2-and-muse-code-with-persistent-async-background-agents)
- [orcarouter.ai — Muse Spark 1.2 explained](https://www.orcarouter.ai/blog/meta-muse-spark-1-2-explained)
- [Meta Model API docs: pricing and rate limits (no Muse Spark Open price listed)](https://dev.meta.ai/docs/pricing-rate-limits)
- [Vorp Labs — Muse Spark 1.2 release review](https://vorplabs.com/models/releases/muse-spark-1-2)
- [explainx.ai — Muse Code Beta](https://www.explainx.ai/blog/meta-muse-code-coding-agent-muse-spark-1-2-launch-august-2026)

Canonical: https://neomanex.com/models/muse-spark-1-2