Benchmark Almanac

ViBench 2.0

GPT-6.1 Sol 64.7% leads since Sep 2026 Not in model cards

Agents build 17 real web apps from empty projects one feature at a time from short specs. A grading agent uses each app in a browser and follows test plans.

8 apps are public and 9 held out, with 241 feature stages and 544 test plans. The paper (ACM CAIS 2026) describes the first version.

Made by ViBench team (Replit and collaborators) Released May 2026 Building whole projects Public tasks plus a private test set Last model added Sep 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
ViBench/vibench-public 44 stars Paper Leaderboard vibench.ai
9
Models scored on its own leaderboard

GPT-6.1 Sol leads with 64.7%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

  1. 64.7 GPT-6.1 Sol
  2. 60.3 GPT-6 Astra
  3. 48.5 Claude Opus 5.5
  4. 19.1 Claude Sonnet 5.5
  5. 1.50 Grok 4.7
  6. 0.0 Kimi K3
  7. 0.0 GLM-5.3
  8. 0.0 Gemini 3.8 Flash
  9. 0.0 DeepSeek V4.1 Flash
  1. 1 GPT-6.1 Sol 64.7%
  2. 2 GPT-6 Astra 60.3%
  3. 3 Claude Opus 5.5 48.5%
  4. 4 Claude Sonnet 5.5 19.1%
  5. 5 Grok 4.7 1.50%
  6. 6 Kimi K3 0%
  7. 7 GLM-5.3 0%
  8. 8 Gemini 3.8 Flash 0%
  9. 9 DeepSeek V4.1 Flash 0%
#ModelScoreSettingDateSource
1 GPT-6.1 Sol 64.7% Sep 2026 Official leaderboard
2 GPT-6 Astra 60.3% Sep 2026 Official leaderboard
3 Claude Opus 5.5 48.5% Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 19.1% Sep 2026 Official leaderboard
5 Grok 4.7 1.50% Sep 2026 Official leaderboard
6 Kimi K3 0% Jul 2026 Official leaderboard
7 GLM-5.3 0% Aug 2026 Official leaderboard
8 Gemini 3.8 Flash 0% Sep 2026 Official leaderboard
9 DeepSeek V4.1 Flash 0% Sep 2026 Official leaderboard

The best score, 64.7%, 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.

DeepSeek Anthropic OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% $1.00 $2.00 $5.00 $10 $20 $50 $100 Cost per task, log scale GPT-6.1 Sol Claude Sonnet 5.5 DeepSeek V4.1 Flash 0% 20% 40% 60% 80% $1.00 $10 $100 Cost per task, log scale GPT-6.1 Sol Claude Sonnet 5.5
ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6.1 Sol 64.7% $12
GPT-6 Astra 60.3% $90
Claude Opus 5.5 48.5% $23
Claude Sonnet 5.5 19.1% $4.36
Grok 4.7 1.50% $64
Kimi K3 0% $33
GLM-5.3 0% $28
Gemini 3.8 Flash 0% $27
DeepSeek V4.1 Flash 0% $2.54

3 models have held number one since its release in May 2026

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

Moonshot AI OpenAI
Own leaderboard 3 leaders
GPT-6.1 Sol 64.7%

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-6.1 Sol GPT-6 Astra Kimi K3 0% 20% 40% 60% 80% 100% Near the ceiling (85%) GPT-6.1 Sol Kimi K3

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Coding

All 243 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%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Opus 5.5 93.9%
  • 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%
  • 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