Pick this when
Local agents on one GPU. Permissive license. Not a cloud flagship.
Skip when
You expected Llama-at-the-frontier. This is a small open agent.
Meta shipping small open models again after going closed with Spark. Useful on a desk, not a leaderboard.
Price
- Input / 1M
- $0.08
- Output / 1M
- $0.24
- Measured task
- $0.04
30B. Fits a consumer GPU. Apache 2.0.
Shape
- Context
- 128K
- Max output
- 16K
- Released
- 2026-08-10
- Tools
- Yes
- Reasoning
- Yes
Scores
Intelligence index38
GPQA Diamond54
Terminal-Bench 2.129
Agentic index18
- Arena
- —
- Speed
- 85/s
- GPQA
- 54.1%
- Term.
- 28.6%
- AA Index. The single number most people mean by “how smart.” It is a blend, so a specialist can lose here and still win the job you care about.
- GPQA. A clean test of hard reasoning. PhD experts sit around 65%. It says little about writing, tools, or taste.
- Term.. Closest public proxy for coding agents that live in a terminal. A high index score with a weak terminal score is a warning.
- Arena. Captures taste and usefulness that unit tests miss. Sample size varies by model, so treat gaps under ~100 Elo as noise.
Meta · Menlo Park. Muse is closed and powers Meta AI across Facebook, Instagram, and WhatsApp. Llama 4 still ships for local work. Glimmer is the small Apache agent.