Benchmark Almanac

Accelerate Bench

Claude Sonnet 5.5 67.4% leads since November 2026
Not in model cards

Coding agents speed up 32 real scientific applications, such as GROMACS and OpenFOAM, on a fixed CPU allocation within 8 hours. A hidden verifier checks correctness, and Score@1 averages the share of runtime saved, with incorrect programs scoring zero.

A correct 2x speedup earns 50% on a task. Paper results use version 1.1, where 16 agents scored 25.5% to 67.4%. 31 tasks are open on Harbor Hub, and the full 32 need approval on Hugging Face. Training on the tasks is prohibited.

Released
November 2026
Kind
Speed and optimization
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Efficiency
Language
English
In and out
Text, Code → Code, Actions
Interaction
Agentic
Environment
Terminal
Items from
Real work and use
Run by
Anyone
In views
Agents
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
16
Models scored on its own leaderboard

Claude Sonnet 5.5 leads with 67.4%

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.

Source: Official Accelerate Bench leaderboard

#ModelScoreSettingDateSource
1 Claude Sonnet 5.5 67.4% high effort; benchmark v1.1, Claude Code Sep 2026 Official leaderboard
2 Claude Opus 5.5 59.4% high effort; benchmark v1.1, Claude Code Sep 2026 Official leaderboard
3 GPT-6.1 Sol 49.2% high effort; benchmark v1.1, Codex Sep 2026 Official leaderboard
4 GPT-6 Astra 47.2% high effort; benchmark v1.1, Codex Sep 2026 Official leaderboard
5 Tencent Hy4 preview 45.2% high effort; benchmark v1.1, Claude Code Aug 2026 Official leaderboard
6 Tencent Hy4 45% high effort; benchmark v1.1, Claude Code Nov 2026 Official leaderboard
7 Kimi K3 42.5% high effort; benchmark v1.1, Claude Code Jul 2026 Official leaderboard
8 DeepSeek V4.1 Flash 1017 40.1% high effort; benchmark v1.1, Claude Code Oct 2026 Official leaderboard
9 GLM-5.3 39.7% high effort; benchmark v1.1, Claude Code Aug 2026 Official leaderboard
10 GLM-5.3-Flash 39.3% high effort; benchmark v1.1, Claude Code Aug 2026 Official leaderboard
11 DeepSeek-V4-Pro 37.9% high effort; benchmark v1.1, Claude Code Apr 2026 Official leaderboard
12 GPT-6.1 Terra 37.9% high effort; benchmark v1.1, Codex Oct 2026 Official leaderboard
13 DeepSeek-V4-Flash 34.1% high effort; benchmark v1.1, Claude Code Apr 2026 Official leaderboard
14 GLM-5.2 34% high effort; benchmark v1.1, Claude Code Jun 2026 Official leaderboard
15 Grok 4.6 33.9% high effort; benchmark v1.1, Claude Code Aug 2026 Official leaderboard
16 Tencent Hy3 25.5% high effort; benchmark v1.1, Claude Code Jul 2026 Official leaderboard

Rankings by part

We read 32 parts from the benchmark’s own leaderboard in 8 groups: molecular simulation & electronic structure, fluids, atmosphere & ocean, transport, plasma & astronomy, bioinformatics & phylogenetics, numerics, tensors & graphs, quantum circuits, machine learning and scientific i/o. Each ranks models on one part of the test. Claude Opus 5.5 leads 15 of them, and 6 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

GROMACS

16 models we track have a score here.

  1. 1 Claude Sonnet 5.5 77.6%
  2. 2 Claude Opus 5.5 77.4%
  3. 3 Tencent Hy3 ↑13 57%
  4. 4 GPT-6 Astra 56.5%
  5. 5 GPT-6.1 Sol ↓2 54.4%
  6. 6 DeepSeek V4.1 Flash 1017 ↑2 54%
  7. 7 DeepSeek-V4-Flash ↑6 53.7%
  8. 8 Tencent Hy4 preview ↓3 53.2%
  9. 9 Tencent Hy4 ↓3 52.8%
  10. 10 GLM-5.3-Flash 52.4%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Own leaderboard 1 leader

Source: Official Accelerate Bench leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Sonnet 5.5 Claude Opus 5.5 GPT-6 Astra Kimi K3 DeepSeek-V4-Pro 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Sonnet 5.5 Claude Opus 5.5

Source: Official Accelerate Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Coding

All 523 in Coding
  • SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
  • LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests. DeepSeek-V4-Pro 93.5%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
  • DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
  • SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases. Seed2.1 Pro 59.8%
  • Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance. DeepSeek V4.1 Flash 3,471

Data audit

All 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (16 entries).
  • Pass Models scored 16 models we track.
  • Pass Newest score The newest is from November 2026, for Tencent Hy4.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 67.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on acceleratebench.com: 16 rows for models we track at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check