AI;DR
AI;DR
← Back to live feed1 story

Thursday, Sep 3, 2026

1

Meta Releases Muse Spark 1.3 Tied With GPT-5.6 Sol at 61 and Cheapest Above 59 on Artificial Analysis Index

Developers can now access Muse Spark 1.3 in Muse Code and the Meta Model API, and OpenCode separately listed the model as free. The generally available xhigh version scored 61 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Sol, and cost $0.55 per task versus $0.95 for Sol. That made it the lowest-cost model above 59 on the index. The max version, in limited preview for partners, scored 62.

Other benchmark figures showed DeepSWE at 75.4, up from 55.0, and Terminal-bench 2.1 at 88.8, tied with GPT-5.6 Sol. An MRCR long-context score of 98.1, up from 55.5, drew pushback from one researcher who said the benchmark can be overfit and that Muse Spark 1.3 did not appear better on other long-context tests.

Sign in to suggest edits

Key sources

  1. SOURCE@jilin_14“The leap from 1.0 to 1.3 is nothing short of a miracle”x.com
  2. SUPPORT@wesroth“from 55.0 to 75.4 on DeepSWE.”x.com
  3. SUPPORT@wesroth“And the limited-preview max version pushes that score to 62.”x.com
  4. SUPPORT@kimmonismus“Meta Muse Spark 1.3 is now free on OpenCode”x.com
  5. SUPPORT@eliebakouch“muse spark 1.3 doesn't seem to be better on other long context benchmarks => very likely benchmaxed on this specific task”x.com
  6. SOURCE@shishirpatil_“Muse Spark 1.3 is rolling out today with frontier performance almost too cheap to meter.”x.com
  7. SUPPORT@arthurmacwaters“impressive performance from muse spark 1.3, especially given it's a fraction of the cost of fable”x.com
  8. SUPPORT@artificialanlys“Meta's Muse Spark 1.3 (max), which is in limited preview for Meta's partners, scores 68 on the Artificial Analysis Coding Agent Index in the Muse Code harness, #2 behind only Claude Opus 5 (xhigh) in Claude Code.”x.com
Markdown