aiexpert
Home / News / Brief
Research · Aug 07, 2026, 07:04 AM · 7 sources

Meta Muse Spark 1.2 reaches 54 on Artificial Analysis Index with coding gains

Meta released Muse Spark 1.2 on August 5, 2026, its third model release in four months, alongside Muse Code, a terminal coding agent shipped in beta. On Artificial Analysis' Intelligence Index, Muse Spark 1.2 scored 54, up from Muse Spark 1.1's 51 and the original Muse Spark's 43 in April—an 11-point gain in four months. The model now ties with SpaceXAI's Grok 4.5 at 54, positioning it just behind the frontier cluster of Claude Opus 5 (61), Claude Fable 5 (60), and GPT-5.6 Sol (59).

The gain concentrated in agentic performance, where Meta's GDPval-AA v2 Elo jumped 260 points to 1631, placing it fifth among benchmarked models. Pricing remains unchanged at $1.25 per million input tokens and $4.25 per million output tokens on the Meta Model API; a new contributor tier offers a 12.5x input and 21.25x output discount in exchange for training rights on user data. Muse Code, the co-trained terminal agent, performs repository-scale software engineering with persistent background agents and parallel subagents. Meta's internal coding benchmarks show Muse Spark 1.2 at 82.9% on Terminal-Bench 2.1 and 59.3% on DeepSWE 1.1—second only to Claude Opus 5 on those vendor-run tests.

For practitioners, the cadence matters: four weeks between major model drops is a pace only Meta maintains now. However, benchmark gains came with increased token usage—input tokens up ~53%, output tokens ~36%—driven by mandatory reasoning at launch. No independent Terminal-Bench verification existed at publication, and Meta's 1.1 came in 3.8 points below its own claim when verified by the benchmark team. A 1M-token context window and multimodal support remain unchanged. Muse Code is available for macOS and Linux, with Windows support not yet announced.

Sources

Everything this brief rests on
  1. 01 Primary source artificialanalysis.ai
  2. 02 artificialanalysis.ai artificialanalysis.ai “Muse Spark 1.2 (xhigh) lands at 54, up 3 points from Muse Spark 1.1 (51) and 11 points from Muse Spark 1.0 (43, April).”
  3. 03 datanorth.ai datanorth.ai “Muse Spark 1.2 scores 82.9% on Terminal-Bench 2.1 at unchanged standard API pricing.”
  4. 04 orcarouter.ai orcarouter.ai “The original Muse Spark scored 43 in April. Muse Spark 1.1 reached 51 on July 9, an eight-point gain in a quarter. Muse Spark 1.2 adds three more in under a month.”
  5. 05 vorplabs.com vorplabs.com “Standard-tier API pricing is unchanged from 1.1 at $1.25 input, $0.15 cached input, and $4.25 output per 1M tokens with a 1,048,576-token context. The real pricing story is the new muse-spark-1.2-contributor tier: $0.10 input and $0.20 output, a discount of 12.5x on input and 21.25x on output.”
  6. 06 artificialanalysis.ai artificialanalysis.ai “Its GDPval-AA v2 Elo rose 260 points to 1631, #5 among all models we have benchmarked”
  7. 07 kingy.ai kingy.ai “Meta's previous model came in 3.8 points below its own published figure when the same benchmark was run and verified by the Terminal-Bench team.”