specs at a glance
| Leaderboard status | Verifying — no confirmed SWE-bench Verified score |
|---|---|
| SWE-bench Verified | — |
| SWE-bench Pro | — |
| Terminal-Bench | — |
| Input price / 1M | Free |
| Output price / 1M | Free |
| Context window | — |
| Open weights | Yes |
| Access | Open weights (Apache 2.0) · Hugging Face · Ollama · LM Studio · Unsloth · llama.cpp · ExecuTorch · MLX · vLLM · SGLang · Together AI · Fireworks · OpenRouter |
| Maker | Meta |
how good is Muse Glimmer at coding?
Muse Glimmer is on our board but deliberately unranked. We rank by SWE-bench Verified, Meta has not published that number, and no independent evaluation has run it, so there is nothing here we could stand behind. It moves into the ranking the moment a confirmed score exists.
Score provenance: Released Aug 10, 2026 with weights on Hugging Face under a plain Apache 2.0 license, Meta's first genuinely permissive open-weight release in the Muse line, which had been API-only until now. Meta published NO SWE-bench Verified score, no SWE-bench Pro and no Terminal-Bench figure: the launch post benchmarks it against Gemma4-31B and Qwen3.6-27B in its own size class and says only that it "performs strongly for its size class".
With no number of any provenance, vendor or independent, there is nothing to rank, so it enters unranked. vals.ai has not evaluated it. Specs that are confirmed: 30B dense, logit-distilled from Muse Spark as teacher, multimodal via a dedicated perception encoder, 100+ languages, OpenClaw scaffold compatible.
Memory: 55GB+ at full bf16, compressed to under 20GB by Meta's K-Quant-17GB build (roughly 4-bit), which is what lets the weights, KV cache, perception encoder and speculative-decoding drafter share a 24GB or 32GB card. Meta reports DFlash speculative decoding at 3.1x on an RTX 5090, 1.8x on an M5 Max and 1.5x on an M4 Max, all vendor-reported.
Separately, Mark Zuckerberg and Alexandr Wang both said an open-weight version of Muse Spark 1.2 (ranked #9 here at 86.6%) follows in the coming weeks; the license for that release has not been stated, and it is a roadmap item until the files land. See /p/meta-muse-glimmer-open-weights-30b/.
Re-checked 2026-09-09: vals.ai's SWE-bench Verified board is unchanged since the September 1 archival and still shows no evaluation for this model; all previously-tracked ranked scores were also re-verified this run and hold within normal variance (largest drift under 0.4pp).
what does Muse Glimmer cost?
Free per 1M input tokens and Free per 1M output — as listed by the maker. Coding workloads are output-heavy, so weight the output rate when budgeting. Run your own volume through the AI API cost calculator for a monthly estimate.
where can you use it?
Available via Open weights (Apache 2.0) · Hugging Face · Ollama · LM Studio · Unsloth · llama.cpp · ExecuTorch · MLX · vLLM · SGLang · Together AI · Fireworks · OpenRouter. Because it ships open weights, you can also self-host it on your own hardware or any inference provider — with the version pinned so the model can't change under you.
Full storyMeta returns to open weights with Muse Glimmer 30B
Ranked on our AI Coding Leaderboard — scores confirmed against primary sources only, updated 2026-09-03.