Benchmark Almanac

Next.js Agent Evals

Claude Fable 5.1 97% leads since Sep 2026 Not in model cards

Coding agents with their models do real Next.js tasks. The board reports success rate with and without an AGENTS.md file, plus duration and cost.

Rerun whenever the evals change, last on 2026-10-30. Top agents reach 97%, so it is near saturation. About 330 GitHub stars.

Made by Vercel Released Oct 2025 Web and app front ends All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
vercel/next-evals-oss 327 stars Leaderboard nextjs.org
16
Entries on its own leaderboard, 14 models we track

5 models share the top score of 97%

Best score per entry 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. 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. 97.0 Claude Fable 5.1
  2. 97.0 Claude Opus 5.5
  3. 97.0 GPT-6 Sol
  4. 97.0 Gemini 4 Argon
  5. 97.0 GPT-6.1 Terra
  6. 94.0 Grok 4.7
  7. 90.0 Pixel Canary
  8. 90.0 Gemini 3.8 Flash
  9. 90.0 GPT-6 Astra
  10. 84.0 Kimi K3
  11. 81.0 Cursor Composer 2.5
  12. 81.0 MiniMax-M3
  13. 81.0 GLM-5.2
  14. 81.0 Claude Sonnet 5
  15. 81.0 GPT-6 Luna
  16. 74.0 Kimi K2.7 Code
  1. 1 Claude Fable 5.1 high 97%
  2. 2 Claude Opus 5.5 high 97%
  3. 3 GPT-6 Sol high 97%
  4. 4 Gemini 4 Argon 97%
  5. 5 GPT-6.1 Terra high 97%
  6. 6 Grok 4.7 94%
  7. 7 Pixel Canary 90%
  8. 8 Gemini 3.8 Flash 90%
  9. 9 GPT-6 Astra high 90%
  10. 10 Kimi K3 84%
  11. 11 Cursor Composer 2.5 81%
  12. 12 MiniMax-M3 81%
  13. 13 GLM-5.2 81%
  14. 14 Claude Sonnet 5 81%
  15. 15 GPT-6 Luna high 81%
  16. 16 Kimi K2.7 Code 74%
#EntryScoreSettingDateSource
1 Claude Fable 5.1 97% high, Claude Code Sep 2026 Official leaderboard
2 Claude Opus 5.5 97% high, Claude Code Sep 2026 Official leaderboard
3 GPT-6 Sol 97% high, Codex Sep 2026 Official leaderboard
4 Gemini 4 Argon 97% OpenCode Sep 2026 Official leaderboard
5 GPT-6.1 Terra 97% high, Codex Oct 2026 Official leaderboard
6 Grok 4.7 94% OpenCode Sep 2026 Official leaderboard
7 Pixel Canary 90% OpenCode Official leaderboard
8 Gemini 3.8 Flash 90% OpenCode Sep 2026 Official leaderboard
9 GPT-6 Astra 90% high, Codex Sep 2026 Official leaderboard
10 Kimi K3 84% OpenCode Jul 2026 Official leaderboard
11 Cursor Composer 2.5 81% Cursor Official leaderboard
12 MiniMax-M3 81% OpenCode Jun 2026 Official leaderboard
13 GLM-5.2 81% OpenCode Jun 2026 Official leaderboard
14 Claude Sonnet 5 81% Claude Code Jun 2026 Official leaderboard
15 GPT-6 Luna 81% high, Codex Sep 2026 Official leaderboard
16 Kimi K2.7 Code 74% OpenCode Jun 2026 Official leaderboard

The best score, 97%, costs $0.08 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.

OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.02 $0.05 $0.10 $0.20 $0.50 $1.00 Cost per task, log scale GPT-6.1 Terra, high GPT-6 Luna, high 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 Cost per task, log scale GPT-6.1 Terra, high GPT-6 Luna, high
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.1 high Claude Code 97% $0.72
Claude Opus 5.5 high Claude Code 97% $0.23
GPT-6 Sol high Codex 97% $0.26
Gemini 4 Argon OpenCode 97% $0.43
GPT-6.1 Terra high Codex 97% $0.08
Grok 4.7 OpenCode 94% $0.18
Gemini 3.8 Flash OpenCode 90% $0.60
GPT-6 Astra high Codex 90% $0.97
Kimi K3 OpenCode 84% $0.28
MiniMax-M3 OpenCode 81% $0.05
GLM-5.2 OpenCode 81% $0.15
Claude Sonnet 5 Claude Code 81% $0.39
GPT-6 Luna high Codex 81% $0.01
Kimi K2.7 Code OpenCode 74% $0.14

3 models have held number one since its release in Oct 2025

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

MiniMax Moonshot AI Anthropic
Own leaderboard Passed 85% in Sep 2026, by Claude Fable 5.1
Claude Fable 5.1 97%

The best score passed 85% in Sep 2026

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% 2026 2027 Near the ceiling (85%) Claude Fable 5.1 Kimi K3 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Fable 5.1 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