Benchmark Almanac

DAYJOB: Data Science

Claude Opus 5.5 37.3% leads since October 2026
Not in model cards

Agents do data science assignments from a short request and a workspace of files. An agent judge applies an expert rubric, and a pass needs 95% of criteria.

Results for 15 models appear only in the Fireworks Specialized Intelligence Index. Surge had not published the tasks or task count by October 30th, 2026.

Made by
Surge AI
Released
October 2026
Kind
Data science and SQL
Level
Expert
Scored
Graded by a model
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
16
Entries on its own leaderboard, 15 models we track

Claude Opus 5.5 leads with 37.3%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

16 entries. Scroll sideways for the rest.

  1. 37.3 Claude Opus 5.5
  2. 35.3 Claude Sonnet 5.5
  3. 26.0 Claude Fable 5.1
  4. 23.3 Gemini 4 Argon
  5. 21.3 GPT-6 Astra
  6. 18.7 GPT-6.1 Sol
  7. 16.7 Muse Wonder 1.0
  8. 7.30 Qwen 3.8 Max
  9. 6.70 Grok 4.7
  10. 6.00 Muse Spark 1.3
  11. 4.70 Kimi K3.1 Pro
  12. 4.00 GLM-5.3
  13. 3.30 Tencent Hy4 preview
  14. 3.30 DeepSeek V4.1 Flash 1017
  15. 1.30 Kimi K3
  16. 0.70 Mistral Large 4
  1. 1 Claude Opus 5.5 adaptive/max 37.3%
  2. 2 Claude Sonnet 5.5 adaptive/max 35.3%
  3. 3 Claude Fable 5.1 adaptive/max 26%
  4. 4 Gemini 4 Argon high 23.3%
  5. 5 GPT-6 Astra max 21.3%
  6. 6 GPT-6.1 Sol max 18.7%
  7. 7 Muse Wonder 1.0 max 16.7%
  8. 8 Qwen 3.8 Max xhigh 7.30%
  9. 9 Grok 4.7 xhigh 6.70%
  10. 10 Muse Spark 1.3 max 6.00%
  11. 11 Kimi K3.1 Pro 4.70%
  12. 12 GLM-5.3 max 4.00%
  13. 13 Tencent Hy4 preview high 3.30%
  14. 14 DeepSeek V4.1 Flash 1017 max 3.30%
  15. 15 Kimi K3 max 1.30%
  16. 16 Mistral Large 4 high 0.70%

Source: Official DAYJOB: Data Science leaderboard

#EntryScoreSettingDateSource
1 Claude Opus 5.5 37.3% adaptive/max Sep 2026 Official leaderboard
2 Claude Sonnet 5.5 35.3% adaptive/max Sep 2026 Official leaderboard
3 Claude Fable 5.1 26% adaptive/max Sep 2026 Official leaderboard
4 Gemini 4 Argon 23.3% high Sep 2026 Official leaderboard
5 GPT-6 Astra 21.3% max Sep 2026 Official leaderboard
6 GPT-6.1 Sol 18.7% max Sep 2026 Official leaderboard
7 Muse Wonder 1.0 16.7% max Nov 2026 Official leaderboard
8 Qwen 3.8 Max 7.30% xhigh Aug 2026 Official leaderboard
9 Grok 4.7 6.70% xhigh Sep 2026 Official leaderboard
10 Muse Spark 1.3 6.00% max Sep 2026 Official leaderboard
11 Kimi K3.1 Pro 4.70% Oct 2026 Official leaderboard
12 GLM-5.3 4.00% max Aug 2026 Official leaderboard
13 Tencent Hy4 preview 3.30% high Aug 2026 Official leaderboard
14 DeepSeek V4.1 Flash 1017 3.30% max Oct 2026 Official leaderboard
15 Kimi K3 1.30% max Jul 2026 Official leaderboard
16 Mistral Large 4 0.70% high Oct 2026 Official leaderboard

Score vs. cost per task

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0% 10% 20% 30% 40% $0$10$20$30 Cost per task ($), lower is better Score Kimi K3 (max): Score 1.3%, cost per task $8.43 Qwen 3.8 Max (xhigh): Score 7.3%, cost per task $9.12 GLM-5.3 (max): Score 4%, cost per task $11.9 Tencent Hy4 preview (high): Score 3.3%, cost per task $3.01 Claude Fable 5.1 (adaptive/max): Score 26%, cost per task $25.5 Muse Spark 1.3 (max): Score 6%, cost per task $4.49 GPT-6 Astra (max): Score 21.3%, cost per task $15.3 Grok 4.7 (xhigh): Score 6.7%, cost per task $8.61 Mistral Large 4 (high): Score 0.7%, cost per task $9.14 Kimi K3.1 Pro: Score 4.7%, cost per task $4.95 Muse Wonder 1.0 (max): Score 16.7%, cost per task $12.8. Not in our model list Claude Opus 5.5 (adaptive/max): Score 37.3%, cost per task $18.9 Claude Sonnet 5.5 (adaptive/max): Score 35.3%, cost per task $16.9 GPT-6.1 Sol (max): Score 18.7%, cost per task $2.68 Gemini 4 Argon (high): Score 23.3%, cost per task $8.19 DeepSeek V4.1 Flash 1017 (max): Score 3.3%, cost per task $0.89 Claude Opus 5.5 (adaptive/max)Claude Sonnet 5.5 (adaptive/max)Gemini 4 Argon (high)GPT-6.1 Sol (max)DeepSeek V4.1 Flash 1017 (max) 0% 10% 20% 30% 40% $0$10$20$30 Cost per task ($), lower is better Score Kimi K3 (max): Score 1.3%, cost per task $8.43 Qwen 3.8 Max (xhigh): Score 7.3%, cost per task $9.12 GLM-5.3 (max): Score 4%, cost per task $11.9 Tencent Hy4 preview (high): Score 3.3%, cost per task $3.01 Claude Fable 5.1 (adaptive/max): Score 26%, cost per task $25.5 Muse Spark 1.3 (max): Score 6%, cost per task $4.49 GPT-6 Astra (max): Score 21.3%, cost per task $15.3 Grok 4.7 (xhigh): Score 6.7%, cost per task $8.61 Mistral Large 4 (high): Score 0.7%, cost per task $9.14 Kimi K3.1 Pro: Score 4.7%, cost per task $4.95 Muse Wonder 1.0 (max): Score 16.7%, cost per task $12.8. Not in our model list Claude Opus 5.5 (adaptive/max): Score 37.3%, cost per task $18.9 Claude Sonnet 5.5 (adaptive/max): Score 35.3%, cost per task $16.9 GPT-6.1 Sol (max): Score 18.7%, cost per task $2.68 Gemini 4 Argon (high): Score 23.3%, cost per task $8.19 DeepSeek V4.1 Flash 1017 (max): Score 3.3%, cost per task $0.89 Claude Opus 5.5 (adaptive/max)Claude Sonnet 5.5 (adaptive/max)Gemini 4 Argon (high)GPT-6.1 Sol (max)DeepSeek V4.1 Flash 1017 (max)
ModelSettingScoreCost per task
Claude Opus 5.5 adaptive/max 37.3% $18.86
Claude Sonnet 5.5 adaptive/max 35.3% $16.89
Claude Fable 5.1 adaptive/max 26% $25.54
Gemini 4 Argon high 23.3% $8.19
GPT-6 Astra max 21.3% $15.34
GPT-6.1 Sol max 18.7% $2.68
Muse Wonder 1.0 max 16.7% $12.83
Qwen 3.8 Max xhigh 7.3% $9.12
Grok 4.7 xhigh 6.7% $8.61
Muse Spark 1.3 max 6% $4.49
Kimi K3.1 Pro 4.7% $4.95
GLM-5.3 max 4% $11.87
DeepSeek V4.1 Flash 1017 max 3.3% $0.89
Tencent Hy4 preview high 3.3% $3.01
Kimi K3 max 1.3% $8.43
Mistral Large 4 high 0.7% $9.14

The best score, 37.3%, costs $19 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Anthropic OpenAI Google DeepMind DeepSeek Cost frontier One model at each effort level
0% 10% 20% 30% 40% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5, adaptive/max Claude Sonnet 5.5, adaptive/max Gemini 4 Argon, high GPT-6.1 Sol, max DeepSeek V4.1 Flash 1017, max 0% 10% 20% 30% 40% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Opus 5.5, adaptive/max Gemini 4 Argon, high

Source: Official DAYJOB: Data Science leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 adaptive/max 37.3% $19
Claude Sonnet 5.5 adaptive/max 35.3% $17
Claude Fable 5.1 adaptive/max 26% $26
Gemini 4 Argon high 23.3% $8.19
GPT-6 Astra max 21.3% $15
GPT-6.1 Sol max 18.7% $2.68
Qwen 3.8 Max xhigh 7.30% $9.12
Grok 4.7 xhigh 6.70% $8.61
Muse Spark 1.3 max 6.00% $4.49
Kimi K3.1 Pro 4.70% $4.95
GLM-5.3 max 4.00% $12
Tencent Hy4 preview high 3.30% $3.01
DeepSeek V4.1 Flash 1017 max 3.30% $0.89
Kimi K3 max 1.30% $8.43
Mistral Large 4 high 0.70% $9.14

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
Claude Opus 5.5 37.3%

Source: Official DAYJOB: Data Science leaderboard

Scores over time

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

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1 Qwen 3.8 Max Kimi K3 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1

Source: Official DAYJOB: Data Science leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Coding

All 282 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