ATLAS-Finance
Agents do 100 investment banking tasks inside 13 simulated firms. They gather facts from email, chat, calendar and files over MCP, then build an Excel model graded against expert rubrics.
Pass means every critical criterion passes with no penalty. Bankers wrote the tasks from their own deal work and need 15 to 30 hours each. Launch article results are v1.0.0, run September 3rd, 2026 to September 10th, 2026.
- Made by
- Handshake AI Research
- Released
- September 2026
- Kind
- Finance
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- All public
- Strongest model tested
- GPT-6 Astra #3 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 12th, 2026
Claude Opus 5 leads with 10.7% 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:
- Claude Fable 5.2
-
Gemini 4 Carbon - Claude Opus 5.5
- GPT-6.1 Sol
- and 12 more
- 1 Claude Opus 5 high 10.7%
- 2 Claude Fable 5.1 high 10.3%
- 3 GPT-6 Astra high 9.67%
- 4 Grok 4.6 high 4.67%
- 5 GPT-5.6 Sol high 4.00%
- 6 Kimi K3 high 3.67%
- 7 Gemini 3.8 Flash high 3.67%
- 8 Gemini 3.7 Flash high 3.33%
- 9 GLM-5.3 high 3.00%
- 10
DeepSeek-V4-Pro high 1.67% - 11 Gemini 3.1 Pro high 0.67%
Source: Official ATLAS-Finance leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 10.7% | high | Jul 2026 | Official leaderboard |
| 2 | Claude Fable 5.1 | 10.3% | high | Sep 2026 | Official leaderboard |
| 3 | GPT-6 Astra | 9.67% | high | Sep 2026 | Official leaderboard |
| 4 | Grok 4.6 | 4.67% | high | Aug 2026 | Official leaderboard |
| 5 | GPT-5.6 Sol | 4.00% | high | Jul 2026 | Official leaderboard |
| 6 | Kimi K3 | 3.67% | high | Jul 2026 | Official leaderboard |
| 7 | Gemini 3.8 Flash | 3.67% | high | Sep 2026 | Official leaderboard |
| 8 | Gemini 3.7 Flash | 3.33% | high | Aug 2026 | Official leaderboard |
| 9 | GLM-5.3 | 3.00% | high | Aug 2026 | Official leaderboard |
| 10 | DeepSeek-V4-Pro | 1.67% | high | Apr 2026 | Official leaderboard |
| 11 | Gemini 3.1 Pro | 0.67% | high | Feb 2026 | Official leaderboard |
Pass rate versus task cost
The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.
| Model | Setting | Pass rate | Cost per task |
|---|---|---|---|
| Claude Opus 5 | high | 10.7% | $11.84 |
| Claude Fable 5.1 | high | 10.3% | $16.14 |
| GPT-6 Astra | high | 9.7% | $7.62 |
| Grok 4.6 | high | 4.7% | $5.49 |
| GPT-5.6 Sol | high | 4% | $3.58 |
| Gemini 3.8 Flash | high | 3.7% | $5.94 |
| Kimi K3 | high | 3.7% | $7.95 |
| Gemini 3.7 Flash | high | 3.3% | $4.76 |
| GLM-5.3 | high | 3% | $9.07 |
| DeepSeek-V4-Pro | high | 1.7% | $2.59 |
| Gemini 3.1 Pro | high | 0.7% | $8.1 |
The best score, 10.7%, costs $12 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.
Source: Official ATLAS-Finance leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5 | high | 10.7% | $12 | 73,000 | |
| Claude Fable 5.1 | high | 10.3% | $16 | 76,000 | |
| GPT-6 Astra | high | 9.67% | $7.62 | 27,000 | |
| Grok 4.6 | high | 4.67% | $5.49 | 52,000 | |
| GPT-5.6 Sol | high | 4.00% | $3.58 | 31,000 | |
| Kimi K3 | high | 3.67% | $7.95 | 58,000 | |
| Gemini 3.8 Flash | high | 3.67% | $5.94 | 143,000 | |
| Gemini 3.7 Flash | high | 3.33% | $4.76 | 204,000 | |
| GLM-5.3 | high | 3.00% | $9.07 | 80,000 | |
| DeepSeek-V4-Pro | high | 1.67% | $2.59 | 90,000 | |
| Gemini 3.1 Pro | high | 0.67% | $8.10 | 51,000 |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official ATLAS-Finance leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official ATLAS-Finance 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.
Versions
We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.
Source: Official leaderboards
More 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%