ProgramBench
200 tasks: rebuild a full program from only its compiled binary and docs. Hidden behavioral tests (248k total) must match the original.
Tasks range from jq to FFmpeg and SQLite. The Opus 5.5 system card reports a 100-agent swarm result.
- Released
- May 2026
- Announced on X
- 4.5K views, 45 likes
- Kind
- Building whole projects
- Level
- Research frontier
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Multi-agent
- Environment
- Terminal
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last in a launch post
- November 2026
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
In the authors’ words
Turning ideas into full software projects from scratch has become a popular use case for language models. Agents are being deployed to seed, maintain, and grow codebases over extended periods with minimal human oversight. Such settings require models to make high-level software architecture decisions. However, existing benchmarks measure focused, limited tasks such as fixing a single bug or developing a single, specified feature. We therefore introduce ProgramBench to measure the ability of software engineering agents to develop software holisitically. In ProgramBench, given only a program and its documentation, agents must architect and implement a codebase that matches the reference executable's behavior. End-to-end behavioral tests are generated via agent-driven fuzzing, enabling evaluation without prescribing implementation structure. Our 200 tasks range from compact CLI tools to widely used software such as FFmpeg, SQLite, and the PHP interpreter. We evaluate 9 LMs and find that none fully resolve any task, with the best model passing 95% of tests on only 3% of tasks. Models favor monolithic, single-file implementations that diverge sharply from human-written code.
Claude Opus 5.5 leads with 18.5%
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. Every run is in the next chart.
MiMo-V2.6 Flash leads with 26%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
- 18.5 Claude Opus 5.5
- 9.00 Claude Fable 5.1
- 8.50
Muse Wonder Full
- 5.50
Muse Wonder
- 5.00 GPT-6 Astra
- 4.50 Claude Opus 5
- 2.50
Muse Spark 1.3
- 1.00 GPT-5.6 Sol
- 0.50 GPT-5.5
- 0.50
Gemini 3.6 Flash
- 0.0 GPT-5 mini
- 0.0 Claude Haiku 4.5
- 0.0
Gemini 3 Flash
- 0.0 Claude Opus 4.6
- 0.0 Claude Sonnet 4.6
- 0.0
Gemini 3.1 Pro
- 0.0 GPT-5.4
- 0.0 GPT-5.4 Mini
- 0.0 Claude Opus 4.7
- 0.0
Gemini 3.5 Flash
- 0.0 Claude Opus 4.8
- 0.0 GLM-5.2
- 0.0
Muse Spark 1.1
- 0.0
Muse Spark 1.2
- 0.0
Gemini 3.7 Flash
- 1 Claude Opus 5.5 xhigh 18.5%
- 2 Claude Fable 5.1 xhigh 9.00%
- 3
Muse Wonder Full 8.50%
- 4
Muse Wonder xhigh 5.50%
- 5 GPT-6 Astra max 5.00%
- 6 Claude Opus 5 xhigh 4.50%
- 7
Muse Spark 1.3 max 2.50%
- 8 GPT-5.6 Sol xhigh 1.00%
- 9 GPT-5.5 xhigh 0.50%
- 10
Gemini 3.6 Flash 0.50%
- 11 GPT-5 mini 0%
- 12 Claude Haiku 4.5 0%
- 13
Gemini 3 Flash 0%
- 14 Claude Opus 4.6 0%
- 15 Claude Sonnet 4.6 0%
- 16
Gemini 3.1 Pro 0%
- 17 GPT-5.4 0%
- 18 GPT-5.4 Mini 0%
- 19 Claude Opus 4.7 xhigh 0%
- 20
Gemini 3.5 Flash 0%
- 21 Claude Opus 4.8 xhigh 0%
- 22 GLM-5.2 0%
- 23
Muse Spark 1.1 xhigh 0%
- 24
Muse Spark 1.2 xhigh 0%
- 25
Gemini 3.7 Flash 0%
Source: Official ProgramBench leaderboard
- 26.0 MiMo-V2.6 Flash
- 18.5 Claude Opus 5.5
- 9.00 Claude Fable 5.1
- 8.50
Muse Wonder Full
- 5.50
Muse Wonder
- 5.00 GPT-6 Astra
- 4.50 Claude Opus 5
- 2.50
Muse Spark 1.3
- 1.00 GPT-5.6 Sol
- 0.50 GPT-5.5
- 0.50
Gemini 3.6 Flash
- 0.0 GPT-5 mini
- 0.0 Claude Haiku 4.5
- 0.0
Gemini 3 Flash
- 0.0 Claude Opus 4.6
- 0.0 Claude Sonnet 4.6
- 0.0
Gemini 3.1 Pro
- 0.0 GPT-5.4
- 0.0 GPT-5.4 Mini
- 0.0 Claude Opus 4.7
- 0.0
Gemini 3.5 Flash
- 0.0 Claude Opus 4.8
- 0.0 GLM-5.2
- 0.0
Muse Spark 1.1
- 0.0
Muse Spark 1.2
- 0.0
Gemini 3.7 Flash
- 1 MiMo-V2.6 Flash 26%
- 2 Claude Opus 5.5 xhigh 18.5%
- 3 Claude Fable 5.1 xhigh 9.00%
- 4
Muse Wonder Full 8.50%
- 5
Muse Wonder xhigh 5.50%
- 6 GPT-6 Astra max 5.00%
- 7 Claude Opus 5 xhigh 4.50%
- 8
Muse Spark 1.3 max 2.50%
- 9 GPT-5.6 Sol xhigh 1.00%
- 10 GPT-5.5 xhigh 0.50%
- 11
Gemini 3.6 Flash 0.50%
- 12 GPT-5 mini 0%
- 13 Claude Haiku 4.5 0%
- 14
Gemini 3 Flash 0%
- 15 Claude Opus 4.6 0%
- 16 Claude Sonnet 4.6 0%
- 17
Gemini 3.1 Pro 0%
- 18 GPT-5.4 0%
- 19 GPT-5.4 Mini 0%
- 20 Claude Opus 4.7 xhigh 0%
- 21
Gemini 3.5 Flash 0%
- 22 Claude Opus 4.8 xhigh 0%
- 23 GLM-5.2 0%
- 24
Muse Spark 1.1 xhigh 0%
- 25
Muse Spark 1.2 xhigh 0%
- 26
Gemini 3.7 Flash 0%
Sources: Labs’ model cards and official ProgramBench leaderboard
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 18.5% | xhigh, mini-SWE-agent | Sep 2026 | Sep 2026 | Official leaderboard |
| 2 | Claude Fable 5.1 | 9.00% | xhigh, mini-SWE-agent | Sep 2026 | Sep 2026 | Official leaderboard |
| 3 | | 8.50% | mini-SWE-agent | Oct 2026 | Oct 2026 | Official leaderboard |
| 4 | | 5.50% | xhigh, mini-SWE-agent | Oct 2026 | Sep 2026 | Official leaderboard |
| 5 | GPT-6 Astra | 5.00% | max, mini-SWE-agent | Sep 2026 | Sep 2026 | Official leaderboard |
| 6 | Claude Opus 5 | 4.50% | xhigh, mini-SWE-agent | Jul 2026 | Jul 2026 | Official leaderboard |
| 7 | | 2.50% | max, mini-SWE-agent | Sep 2026 | Sep 2026 | Official leaderboard |
| 8 | GPT-5.6 Sol | 1.00% | xhigh, mini-SWE-agent | Jul 2026 | Jul 2026 | Official leaderboard |
| 9 | GPT-5.5 | 0.50% | xhigh, mini-SWE-agent | Apr 2026 | Apr 2026 | Official leaderboard |
| 10 | | 0.50% | mini-SWE-agent | Jul 2026 | Jul 2026 | Official leaderboard |
| 11 | GPT-5 mini | 0% | mini-SWE-agent | Aug 2025 | Aug 2025 | Official leaderboard |
| 12 | Claude Haiku 4.5 | 0% | mini-SWE-agent | Oct 2025 | Oct 2025 | Official leaderboard |
| 13 | | 0% | mini-SWE-agent | Dec 2025 | Dec 2025 | Official leaderboard |
| 14 | Claude Opus 4.6 | 0% | mini-SWE-agent | Feb 2026 | Feb 2026 | Official leaderboard |
| 15 | Claude Sonnet 4.6 | 0% | mini-SWE-agent | Feb 2026 | Feb 2026 | Official leaderboard |
| 16 | | 0% | mini-SWE-agent | Feb 2026 | Feb 2026 | Official leaderboard |
| 17 | GPT-5.4 | 0% | mini-SWE-agent | Mar 2026 | Mar 2026 | Official leaderboard |
| 18 | GPT-5.4 Mini | 0% | mini-SWE-agent | Mar 2026 | Mar 2026 | Official leaderboard |
| 19 | Claude Opus 4.7 | 0% | xhigh, mini-SWE-agent | Apr 2026 | Apr 2026 | Official leaderboard |
| 20 | | 0% | mini-SWE-agent | May 2026 | May 2026 | Official leaderboard |
| 21 | Claude Opus 4.8 | 0% | xhigh, mini-SWE-agent | May 2026 | May 2026 | Official leaderboard |
| 22 | GLM-5.2 | 0% | mini-SWE-agent | Jun 2026 | Jun 2026 | Official leaderboard |
| 23 | | 0% | xhigh, mini-SWE-agent | Jul 2026 | Jul 2026 | Official leaderboard |
| 24 | | 0% | xhigh, mini-SWE-agent | Aug 2026 | Aug 2026 | Official leaderboard |
| 25 | | 0% | mini-SWE-agent | Aug 2026 | Aug 2026 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | MiMo-V2.6 Flash | 26% | Sep 2026 | Lab model card | |
| 2 | Claude Opus 5.5 | 18.5% | xhigh, mini-SWE-agent | Sep 2026 | Official leaderboard |
| 3 | Claude Fable 5.1 | 9.00% | xhigh, mini-SWE-agent | Sep 2026 | Official leaderboard |
| 4 | 8.50% | mini-SWE-agent | Oct 2026 | Official leaderboard | |
| 5 | 5.50% | xhigh, mini-SWE-agent | Oct 2026 | Official leaderboard | |
| 6 | GPT-6 Astra | 5.00% | max, mini-SWE-agent | Sep 2026 | Official leaderboard |
| 7 | Claude Opus 5 | 4.50% | xhigh, mini-SWE-agent | Jul 2026 | Official leaderboard |
| 8 | 2.50% | max, mini-SWE-agent | Sep 2026 | Official leaderboard | |
| 9 | GPT-5.6 Sol | 1.00% | xhigh, mini-SWE-agent | Jul 2026 | Official leaderboard |
| 10 | GPT-5.5 | 0.50% | xhigh, mini-SWE-agent | Apr 2026 | Official leaderboard |
| 11 | 0.50% | mini-SWE-agent | Jul 2026 | Official leaderboard | |
| 12 | GPT-5 mini | 0% | mini-SWE-agent | Aug 2025 | Official leaderboard |
| 13 | Claude Haiku 4.5 | 0% | mini-SWE-agent | Oct 2025 | Official leaderboard |
| 14 | 0% | mini-SWE-agent | Dec 2025 | Official leaderboard | |
| 15 | Claude Opus 4.6 | 0% | mini-SWE-agent | Feb 2026 | Official leaderboard |
| 16 | Claude Sonnet 4.6 | 0% | mini-SWE-agent | Feb 2026 | Official leaderboard |
| 17 | 0% | mini-SWE-agent | Feb 2026 | Official leaderboard | |
| 18 | GPT-5.4 | 0% | mini-SWE-agent | Mar 2026 | Official leaderboard |
| 19 | GPT-5.4 Mini | 0% | mini-SWE-agent | Mar 2026 | Official leaderboard |
| 20 | Claude Opus 4.7 | 0% | xhigh, mini-SWE-agent | Apr 2026 | Official leaderboard |
| 21 | 0% | mini-SWE-agent | May 2026 | Official leaderboard | |
| 22 | Claude Opus 4.8 | 0% | xhigh, mini-SWE-agent | May 2026 | Official leaderboard |
| 23 | GLM-5.2 | 0% | mini-SWE-agent | Jun 2026 | Official leaderboard |
| 24 | 0% | xhigh, mini-SWE-agent | Jul 2026 | Official leaderboard | |
| 25 | 0% | xhigh, mini-SWE-agent | Aug 2026 | Official leaderboard | |
| 26 | 0% | mini-SWE-agent | Aug 2026 | Official leaderboard |
Labs report scores for 1 model in their cards. Switch the scores above to rank them together.
Other ways to score it
The benchmark’s own leaderboard also reports Almost solved and Tests passed, other ways to score the whole test. Labs sometimes quote these instead of the headline. Claude Opus 5.5 leads Almost solved and Muse Wonder Full leads Tests passed. Arrows show how far a model moved from its overall rank among the models we track.
Almost solved
The share of tasks whose rebuilt program passes at least 95% of the hidden tests, the board's Almost column. Most labs quote it. 25 models we track have a score here.
- 1 Claude Opus 5.5 70.5%
- 2
Muse Wonder Full ↑1 60%
- 3 Claude Fable 5.1 ↓1 55.5%
- 4 GPT-6 Astra ↑1 52.5%
- 5
Muse Wonder ↓1 50.5%
- 6 Claude Opus 5 37%
- 7
Muse Spark 1.3 25%
- 8 Claude Opus 4.8 ↑13 16.5%
- 9 GPT-5.6 Sol ↓1 15.5%
- 10 GPT-5.5 ↓1 13.5%
Tests passed
The share of hidden behavioral tests the rebuilt program passes, averaged over tasks. The board calls it an unofficial metric. Anthropic, ByteDance and xAI quote it. 25 models we track have a score here.
- 1
Muse Wonder Full ↑2 88.5%
- 2 Claude Opus 5.5 ↓1 87.7%
- 3 GPT-6 Astra ↑2 86.9%
- 4
Muse Wonder 86.5%
- 5 Claude Fable 5.1 ↓3 82.5%
- 6 Claude Opus 5 74.7%
- 7 Claude Opus 4.8 ↑14 70.9%
- 8
Muse Spark 1.3 ↓1 70.8%
- 9 GPT-5.6 Sol ↓1 69.9%
- 10 GPT-5.5 ↓1 69.5%
Pareto
The board's Pareto view plots the share of hidden tests passed, which it calls an unofficial metric, against cost, turns, tokens or date. This is its Turns view. 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 | % Test Passed | Turns per task |
|---|---|---|---|
| mini-SWE-agent | 88.5% | 395 | |
| Claude Opus 5.5 | xhigh, mini-SWE-agent | 87.7% | 304 |
| GPT-6 Astra | max, mini-SWE-agent | 86.9% | 57 |
| xhigh, mini-SWE-agent | 86.5% | 363 | |
| Claude Fable 5.1 | xhigh, mini-SWE-agent | 82.5% | 260 |
| Claude Opus 5 | xhigh, mini-SWE-agent | 74.7% | 257 |
| Claude Opus 4.8 | xhigh, mini-SWE-agent | 70.9% | 145 |
| max, mini-SWE-agent | 70.8% | 325 | |
| GPT-5.6 Sol | xhigh, mini-SWE-agent | 69.9% | 42 |
| GPT-5.5 | xhigh, mini-SWE-agent | 69.5% | 82 |
| xhigh, mini-SWE-agent | 68.6% | 253 | |
| GPT-5.5 | high, mini-SWE-agent | 66.5% | 41 |
| GLM-5.2 | mini-SWE-agent | 64.6% | 222 |
| mini-SWE-agent | 61.2% | 133 | |
| GPT-5.6 Sol | mini-SWE-agent | 57.8% | 15 |
| xhigh, mini-SWE-agent | 57.2% | 183 | |
| GPT-5.5 | mini-SWE-agent | 56.6% | 19 |
| mini-SWE-agent | 55.7% | 150 | |
| Claude Opus 4.7 | xhigh, mini-SWE-agent | 55.1% | 159 |
| mini-SWE-agent | 53.6% | 166 | |
| Claude Opus 4.6 | mini-SWE-agent | 52.1% | 260 |
| Claude Opus 4.7 | mini-SWE-agent | 50.9% | 93 |
| Claude Sonnet 4.6 | mini-SWE-agent | 47.5% | 472 |
| xhigh, mini-SWE-agent | 47% | 193 | |
| GPT-5.4 | mini-SWE-agent | 37.7% | 16 |
| mini-SWE-agent | 36.4% | 94 | |
| mini-SWE-agent | 31.8% | 85 | |
| Claude Haiku 4.5 | mini-SWE-agent | 30% | 124 |
| GPT-5.4 Mini | mini-SWE-agent | 16.4% | 18 |
| GPT-5 mini | mini-SWE-agent | 16% | 15 |
The best score, 18.5%, costs $36 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 ProgramBench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5.5 | xhigh | mini-SWE-agent | 18.5% | $36 | 434,511 |
| Claude Fable 5.1 | xhigh | mini-SWE-agent | 9.00% | $57 | 445,587 |
| mini-SWE-agent | 8.50% | 352,279 | |||
| xhigh | mini-SWE-agent | 5.50% | $37 | 345,971 | |
| GPT-6 Astra | max | mini-SWE-agent | 5.00% | $11 | 91,595 |
| Claude Opus 5 | xhigh | mini-SWE-agent | 4.50% | $51 | 357,785 |
| max | mini-SWE-agent | 2.50% | $6.46 | 387,723 | |
| GPT-5.6 Sol | xhigh | mini-SWE-agent | 1.00% | $6.08 | 56,179 |
| xhigh | mini-SWE-agent | 1.00% | $2.02 | 216,872 | |
| GPT-5.5 | xhigh | mini-SWE-agent | 0.50% | $8.85 | 51,624 |
| GPT-5.5 | high | mini-SWE-agent | 0.50% | $3.65 | 27,379 |
| GPT-5.6 Sol | mini-SWE-agent | 0.50% | $1.00 | 13,084 | |
| mini-SWE-agent | 0.50% | $4.83 | 107,792 | ||
| GPT-5 mini | mini-SWE-agent | 0% | $0.03 | 7,146 | |
| Claude Haiku 4.5 | mini-SWE-agent | 0% | $0.80 | 47,506 | |
| mini-SWE-agent | 0% | $0.30 | 22,561 | ||
| Claude Opus 4.6 | mini-SWE-agent | 0% | $11 | 89,147 | |
| Claude Sonnet 4.6 | mini-SWE-agent | 0% | $27 | 190,227 | |
| mini-SWE-agent | 0% | $1.51 | 18,071 | ||
| GPT-5.4 | mini-SWE-agent | 0% | $0.33 | 7,408 | |
| GPT-5.4 Mini | mini-SWE-agent | 0% | $0.04 | 3,147 | |
| Claude Opus 4.7 | xhigh | mini-SWE-agent | 0% | $11 | 97,205 |
| Claude Opus 4.7 | mini-SWE-agent | 0% | $3.81 | 41,669 | |
| GPT-5.5 | mini-SWE-agent | 0% | $1.21 | 11,928 | |
| mini-SWE-agent | 0% | $5.60 | 132,679 | ||
| Claude Opus 4.8 | xhigh | mini-SWE-agent | 0% | $21 | 270,298 |
| GLM-5.2 | mini-SWE-agent | 0% | $25 | 237,624 | |
| xhigh | mini-SWE-agent | 0% | $0.73 | 357,115 | |
| xhigh | mini-SWE-agent | 0% | $1.48 | 329,742 | |
| mini-SWE-agent | 0% | $2.04 | 74,885 |
5 models have held number one on the official leaderboard since its release in May 2026 5 models have held number one with claims included since its release in May 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official ProgramBench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official ProgramBench leaderboard and labs’ model cards
10 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Nov 15, 2026 | 31.5 | Black-box software replication, ProgramBench | |
| Nov 4, 2026 | 30 | ProgramBench (almost solved) * | |
| Nov 2, 2026 | 25.5 | Program Bench (Almost Resolved) | |
| Oct 24, 2026 | Kimi K3.1 Pro | 31.5 | Program Bench (binary score, >=95% counts as solved) |
| Oct 17, 2026 | ProgramBench | ||
| Oct 12, 2026 | ProgramBench | ||
| Oct 12, 2026 | ProgramBench | ||
| Sep 22, 2026 | MiMo-V2.6 Pro | ProgramBench | |
| Sep 21, 2026 | MiMo-V2.6 Flash | 26 | ProgramBench |
| Sep 20, 2026 | 80.5 | ProgramBench (Pass Rate) | |
| Sep 10, 2026 | 20.3 | ProgramBench (Almost@1) | |
| Sep 2, 2026 | 28 | Black-box software replication, ProgramBench (Almost Solved) | |
| Aug 28, 2026 | ProgramBench | ||
| Aug 14, 2026 | GLM-5.3 | 19 | ProgramBench (Almost Solved) |
| Jul 16, 2026 | Kimi K3 | ProgramBench | |
| Jun 23, 2026 | 50.3 | ProgramBench (resolved / almost resolved / average pass rate) 0/1/50.3 | |
| Jun 16, 2026 | GLM-5.2 | ProgramBench | |
| Jun 12, 2026 | Kimi K2.7 Code | Program Bench |
15 system cards report it, from 7 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 9, 2026 | Claude Fable 5.2 system card | 85.6 | ProgramBench hidden-test pass rate on 166 golden tasks, single agent, mini-swe-agent harness, no time limit, 1M-token window. Showing remaining context raises the score to 91.6% (Section 8.10.1) | Capabilities |
| Oct 12, 2026 | | 50 | ProgramBench, rollouts passing at least 95% of tests, mini-swe-agent, Muse Wonder (max) (capabilities table image, page 3) | Capabilities |
| Oct 12, 2026 | | 59 | ProgramBench, rollouts passing at least 95% of tests, mini-swe-agent, Muse Wonder Full (max) (capabilities table image, page 3) | Capabilities |
| Oct 7, 2026 | Claude Haiku 5.5 system card | 82 | ProgramBench, hidden test pass rate on 166 of 200 tasks, mini-swe-agent without time limit (8.7.1) | Capabilities |
| Sep 28, 2026 | Claude Sonnet 5.5 system card | 79.7 | ProgramBench hidden test pass rate on 166 tasks, single agent, mini-swe-agent harness (8.10.1) | Capabilities |
| Sep 22, 2026 | MiMo-V2.6 Pro technical report | ProgramBench (Table 3) | Capabilities | |
| Sep 22, 2026 | Claude Opus 5.5 system card | 91.2 | ProgramBench hidden test pass rate on 166 of 200 tasks, single agent, mini-swe-agent harness, no time limit (8.10.1) | Capabilities |
| Sep 10, 2026 | | 20.3 | ProgramBench (Almost@1) (Table 3). Figure 10 text: 30.04 multi-agent and 20.39 single-agent at 8h on a 172-task high-confidence subset | Capabilities |
| Sep 1, 2026 | Claude Fable 5.1 system card | 87.6 | ProgramBench hidden test pass rate on 166 tasks, single agent, mini-swe-agent harness, no time limit (8.11.1) | Capabilities |
| Jul 27, 2026 | Kimi K3 technical report | ProgramBench (Table 2) | Capabilities | |
| Jul 24, 2026 | Claude Opus 5 system card | 93 | ProgramBench hidden test pass rate on 166 of 200 tasks after five 1M-token episodes, single agent (83% after the first episode) (8.9.1) | Capabilities |
| Jul 14, 2026 | Grok 4.5 model card | 57.2 | ProgramBench, tests-passed rate, high (2.8) | Capabilities |
| Jun 30, 2026 | Claude Sonnet 5 system card | ProgramBench, hidden test pass rate on 166 of 200 tasks across 5 episodes, given only as a range 76 to 86%, no single value (8.8) | Capabilities | |
| Jun 23, 2026 | | 50.3 | ProgramBench, resolved 0 / almost resolved 1 / average pass rate 50.3 (Table 5) | Capabilities |
| May 28, 2026 | Claude Opus 4.8 system card | 79 | ProgramBench hidden test pass rate on 166 of 200 tasks, one episode of up to 1M tokens, near-final snapshot. Text gives 79 to 88% across 1 to 5 episodes (8.5, Figure 8.5.A) | Capabilities |
The lab's number comes out ahead in 4 of 6 pairs that differ
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.
| Model | Lab says | Board says | Gap |
|---|---|---|---|
| Claude Opus 5 Tests passed | 93% ProgramBench hidden test pass rate on 166 of 200 tasks after five 1M-token episodes, single agent (83% after the first episode) (8.9.1) | 74.7% Leaderboard, xhigh, mini-SWE-agent | +18 pts |
| Claude Opus 4.8 Tests passed | 79% ProgramBench hidden test pass rate on 166 of 200 tasks, one episode of up to 1M tokens, near-final snapshot. Text gives 79 to 88% across 1 to 5 episodes (8.5, Figure 8.5.A) | 70.9% Leaderboard, xhigh, mini-SWE-agent | +8.1 pts |
| Claude Fable 5.1 Tests passed | 87.6% ProgramBench hidden test pass rate on 166 tasks, single agent, mini-swe-agent harness, no time limit (8.11.1) | 82.5% Leaderboard, xhigh, mini-SWE-agent | +5.1 pts |
| Claude Opus 5.5 Tests passed | 91.2% ProgramBench hidden test pass rate on 166 of 200 tasks, single agent, mini-swe-agent harness, no time limit (8.10.1) | 87.7% Leaderboard, xhigh, mini-SWE-agent | +3.5 pts |
| | 59% ProgramBench, rollouts passing at least 95% of tests, mini-swe-agent, Muse Wonder Full (max) (capabilities table image, page 3) | 60% Leaderboard, mini-SWE-agent | −1.0 pts |
| | 50% ProgramBench, rollouts passing at least 95% of tests, mini-swe-agent, Muse Wonder (max) (capabilities table image, page 3) | 50.5% Leaderboard, xhigh, mini-SWE-agent | −0.5 pts |
More 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
11 of 12 checks pass, 1 warning. Look first at the Lab claims check. 1 of 6 pairs differs by more than 10 points, the largest for Claude Opus 5 (+18 pts). 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 (25 entries), launch posts (10 scores) and system cards (12 scores).
- Pass Models scored 26 models we track.
- Pass Newest score The newest is from November 2026, for Qwen4-Max.
- Pass Frontier models 3 of the top 10 have a score, led by GPT-6 Astra (#4).
- Pass Headroom The best is 26%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on programbench.com: 30 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 warning 1 pass
- Warning Lab claims 1 of 6 pairs differs by more than 10 points, the largest for Claude Opus 5 (+18 pts).
- Pass Known issues No published audit found problems.