Benchmark Almanac

SheetBench

GPT-6 Astra 64.7% leads since September 2026
Not in model cards

Agents solve finance and accounting spreadsheet tasks through a browser in a fresh Google Sheet. The score is the share of expected answer cells that match exactly.

50 tasks with up to 200 model turns per attempt. 8 models in the Fireworks index. Release date not confirmed on a HUD page.

Made by
HUD
Kind
Finance
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
8
Models scored on its own leaderboard

GPT-6 Astra leads with 64.7% among models released by September 2026

Best score per model 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.

The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 64.7 GPT-6 Astra
  2. 62.5 Claude Opus 5.5
  3. 60.5 GPT-6.1 Sol
  4. 59.3 Kimi K3
  5. 58.4 DeepSeek V4.1 Flash
  6. 45.6 Gemini 3.8 Flash
  7. 42.8 Muse Spark 1.3
  8. 14.3 Grok 4.7
  1. 1 GPT-6 Astra 64.7%
  2. 2 Claude Opus 5.5 62.5%
  3. 3 GPT-6.1 Sol 60.5%
  4. 4 Kimi K3 59.3%
  5. 5 DeepSeek V4.1 Flash 58.4%
  6. 6 Gemini 3.8 Flash 45.6%
  7. 7 Muse Spark 1.3 42.8%
  8. 8 Grok 4.7 14.3%

Source: Official SheetBench leaderboard

#ModelScoreSettingDateSource
1 GPT-6 Astra 64.7% Sep 2026 Official leaderboard
2 Claude Opus 5.5 62.5% Sep 2026 Official leaderboard
3 GPT-6.1 Sol 60.5% Sep 2026 Official leaderboard
4 Kimi K3 59.3% Jul 2026 Official leaderboard
5 DeepSeek V4.1 Flash 58.4% Sep 2026 Official leaderboard
6 Gemini 3.8 Flash 45.6% Sep 2026 Official leaderboard
7 Muse Spark 1.3 42.8% Sep 2026 Official leaderboard
8 Grok 4.7 14.3% Sep 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% 20% 40% 60% 80% $0$2$4$6 Cost per task ($), lower is better Score Kimi K3: Score 59.3%, cost per task $1.62 Gemini 3.8 Flash: Score 45.6%, cost per task $1.9 Muse Spark 1.3: Score 42.8%, cost per task $5.92 DeepSeek V4.1 Flash: Score 58.4%, cost per task $0.45 Grok 4.7: Score 14.3%, cost per task $5.56 GPT-6 Astra: Score 64.7%, cost per task $2.63 Claude Opus 5.5: Score 62.5%, cost per task $1.78 GPT-6.1 Sol: Score 60.5%, cost per task $0.22 GPT-6 AstraClaude Opus 5.5GPT-6.1 Sol 0% 20% 40% 60% 80% $0$2$4$6 Cost per task ($), lower is better Score Kimi K3: Score 59.3%, cost per task $1.62 Gemini 3.8 Flash: Score 45.6%, cost per task $1.9 Muse Spark 1.3: Score 42.8%, cost per task $5.92 DeepSeek V4.1 Flash: Score 58.4%, cost per task $0.45 Grok 4.7: Score 14.3%, cost per task $5.56 GPT-6 Astra: Score 64.7%, cost per task $2.63 Claude Opus 5.5: Score 62.5%, cost per task $1.78 GPT-6.1 Sol: Score 60.5%, cost per task $0.22 GPT-6 AstraClaude Opus 5.5GPT-6.1 Sol
ModelSettingScoreCost per task
GPT-6 Astra 64.7% $2.63
Claude Opus 5.5 62.5% $1.78
GPT-6.1 Sol 60.5% $0.22
Kimi K3 59.3% $1.62
DeepSeek V4.1 Flash 58.4% $0.45
Gemini 3.8 Flash 45.6% $1.9
Muse Spark 1.3 42.8% $5.92
Grok 4.7 14.3% $5.56

The best score, 64.7%, costs $2.63 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol comes within 4.2 points of the best for $0.22 a task, 12 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Anthropic Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.10 $0.20 $0.50 $1.00 $2.00 $5.00 $10 Cost per task, log scale GPT-6 Astra Claude Opus 5.5 GPT-6.1 Sol 0% 20% 40% 60% 80% $0.10 $1.00 $10 Cost per task, log scale GPT-6 Astra GPT-6.1 Sol

Source: Official SheetBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra 64.7% $2.63
Claude Opus 5.5 62.5% $1.78
GPT-6.1 Sol 60.5% $0.22
Kimi K3 59.3% $1.62
DeepSeek V4.1 Flash 58.4% $0.45
Gemini 3.8 Flash 45.6% $1.90
Muse Spark 1.3 42.8% $5.92
Grok 4.7 14.3% $5.56

2 models have held number one on the official leaderboard since 2023

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

Moonshot AI OpenAI
Own leaderboard Not tested since Sep 2026
GPT-6 Astra 64.7%

Source: Official SheetBench 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% Near the ceiling (85%) GPT-6 Astra Kimi K3 0% 20% 40% 60% 80% 100% Near the ceiling (85%) GPT-6 Astra

Source: Official SheetBench 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 Law, finance and medicine

All 179 in Law, finance and medicine
  • MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy. Qwen 3.8 Max 80.4%
  • HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
  • HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy. Ling-3.1-flash 81.7%
  • KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items. Solar Pro 4 77.5%
  • KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. Solar Pro 4 93.2%