Benchmark Almanac

FrontierCode 1.1

Claude Fable 5 53.5% leads since July 2026
GPT-6 Astra 64.5% leads since September 2026
In use

Real-repo coding tasks graded for mergeability: correctness, regressions, lint, tests, scope and code quality. Failing any blocker scores zero.

Re-graded the 1.0 tasks: runs flagged for unfair internet use score zero, and over-strict checks were relaxed.

Made by
Cognition
Released
July 2026
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 12th, 2026
8
Models scored on its own leaderboard
10
Launch posts list it, from 5 labs
7
System cards report it, from 3 labs

Claude Fable 5 leads with 53.5% among models released by June 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.

GPT-6 Astra leads with 64.5%, 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.

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

  1. 53.5 Claude Fable 5
  2. 46.5 Claude Opus 4.8
  3. 43.0 GPT-5.5
  4. 42.7 Claude Sonnet 5
  5. 38.5 Claude Opus 4.7
  6. 27.0 GPT-5.4 Mini
  7. 26.9 Claude Opus 4.6
  8. 24.3 Claude Sonnet 4.6
  1. 1 Claude Fable 5 xhigh 53.5%
  2. 2 Claude Opus 4.8 max 46.5%
  3. 3 GPT-5.5 xhigh 43%
  4. 4 Claude Sonnet 5 xhigh 42.7%
  5. 5 Claude Opus 4.7 max 38.5%
  6. 6 GPT-5.4 Mini xhigh 27%
  7. 7 Claude Opus 4.6 max 26.9%
  8. 8 Claude Sonnet 4.6 max 24.3%

Source: Official FrontierCode 1.1 leaderboard

21 models. Scroll sideways for the rest.

  1. 64.5 GPT-6 Astra
  2. 61.3 Grok 4.6
  3. 54.4 Claude Opus 5.5
  4. 53.5 Claude Fable 5
  5. 53.4 Claude Opus 5
  6. 52.1 Claude Sonnet 5.5
  7. 50.9 Claude Fable 5.1
  8. 50.0 SWE-2
  9. 49.3 GPT-6 Sol
  10. 46.5 Claude Opus 4.8
  11. 46.4 Claude Haiku 5.5
  12. 45.5 SWE-2 Flash
  13. 43.6 Gemini 3.7 Flash
  14. 43.0 GPT-5.5
  15. 42.7 Claude Sonnet 5
  16. 42.4 GPT-6 Luna
  17. 42.3 SWE-1.7
  18. 38.5 Claude Opus 4.7
  19. 27.0 GPT-5.4 Mini
  20. 26.9 Claude Opus 4.6
  21. 24.3 Claude Sonnet 4.6
  1. 1 GPT-6 Astra 64.5%
  2. 2 Grok 4.6 61.3%
  3. 3 Claude Opus 5.5 54.4%
  4. 4 Claude Fable 5 xhigh 53.5%
  5. 5 Claude Opus 5 53.4%
  6. 6 Claude Sonnet 5.5 52.1%
  7. 7 Claude Fable 5.1 50.9%
  8. 8 SWE-2 50%
  9. 9 GPT-6 Sol 49.3%
  10. 10 Claude Opus 4.8 max 46.5%
  11. 11 Claude Haiku 5.5 46.4%
  12. 12 SWE-2 Flash 45.5%
  13. 13 Gemini 3.7 Flash 43.6%
  14. 14 GPT-5.5 xhigh 43%
  15. 15 Claude Sonnet 5 xhigh 42.7%
  16. 16 GPT-6 Luna 42.4%
  17. 17 SWE-1.7 42.3%
  18. 18 Claude Opus 4.7 max 38.5%
  19. 19 GPT-5.4 Mini xhigh 27%
  20. 20 Claude Opus 4.6 max 26.9%
  21. 21 Claude Sonnet 4.6 max 24.3%

Sources: Labs’ model cards and official FrontierCode 1.1 leaderboard

#ModelScoreSettingDateSource
1 Claude Fable 5 53.5% xhigh, Claude Code Jun 2026 Official leaderboard
2 Claude Opus 4.8 46.5% max, Claude Code May 2026 Official leaderboard
3 GPT-5.5 43% xhigh, Codex Apr 2026 Official leaderboard
4 Claude Sonnet 5 42.7% xhigh, Claude Code Jun 2026 Official leaderboard
5 Claude Opus 4.7 38.5% max, Claude Code Apr 2026 Official leaderboard
6 GPT-5.4 Mini 27% xhigh, Codex Mar 2026 Official leaderboard
7 Claude Opus 4.6 26.9% max, Claude Code Feb 2026 Official leaderboard
8 Claude Sonnet 4.6 24.3% max, Claude Code Feb 2026 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-6 Astra 64.5% Sep 2026 Lab model card
2 Grok 4.6 61.3% Aug 2026 Lab model card
3 Claude Opus 5.5 54.4% Sep 2026 Lab model card
4 Claude Fable 5 53.5% xhigh, Claude Code Jun 2026 Official leaderboard
5 Claude Opus 5 53.4% Jul 2026 Lab model card
6 Claude Sonnet 5.5 52.1% Sep 2026 Lab model card
7 Claude Fable 5.1 50.9% Sep 2026 Lab model card
8 SWE-2 50% Sep 2026 Lab model card
9 GPT-6 Sol 49.3% Sep 2026 Lab model card
10 Claude Opus 4.8 46.5% max, Claude Code May 2026 Official leaderboard
11 Claude Haiku 5.5 46.4% Oct 2026 Lab model card
12 SWE-2 Flash 45.5% Nov 2026 Lab model card
13 Gemini 3.7 Flash 43.6% Aug 2026 Lab model card
14 GPT-5.5 43% xhigh, Codex Apr 2026 Official leaderboard
15 Claude Sonnet 5 42.7% xhigh, Claude Code Jun 2026 Official leaderboard
16 GPT-6 Luna 42.4% Sep 2026 Lab model card
17 SWE-1.7 42.3% Jul 2026 Lab model card
18 Claude Opus 4.7 38.5% max, Claude Code Apr 2026 Official leaderboard
19 GPT-5.4 Mini 27% xhigh, Codex Mar 2026 Official leaderboard
20 Claude Opus 4.6 26.9% max, Claude Code Feb 2026 Official leaderboard
21 Claude Sonnet 4.6 24.3% max, Claude Code Feb 2026 Official leaderboard

Labs report scores for 13 models in their cards. Switch the scores above to rank them together.

The best score, 53.5%, costs $13 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 Anthropic Cost frontier One model at each effort level
0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5, xhigh Claude Fable 5, medium GPT-5.5, xhigh GPT-5.5, high GPT-5.5, medium GPT-5.5, low GPT-5.4 Mini, xhigh 0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5, xhigh Claude Fable 5, medium

Source: Official FrontierCode 1.1 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5 xhigh Claude Code 53.5% $13 58,558
Claude Fable 5 high Claude Code 52.7% $9.48 43,483
Claude Fable 5 max Claude Code 51.6% $19 80,761
Claude Fable 5 medium Claude Code 49.8% $7.15 33,230
Claude Fable 5 low Claude Code 48% $4.92 23,610
Claude Opus 4.8 max Claude Code 46.5% $9.62 95,940
Claude Opus 4.8 xhigh Claude Code 45.5% $6.78 65,742
GPT-5.5 xhigh Codex 43% $4.04 24,992
Claude Sonnet 5 xhigh Claude Code 42.7% $9.06 87,937
Claude Sonnet 5 max Claude Code 42.4% $15 142,621
Claude Opus 4.8 high Claude Code 41% $4.28 41,326
GPT-5.5 high Codex 40.6% $3.12 18,422
Claude Opus 4.8 medium Claude Code 40.5% $3.70 34,907
Claude Sonnet 5 high Claude Code 39.4% $5.49 57,710
Claude Opus 4.7 max Claude Code 38.5% $9.09 49,114
GPT-5.5 medium Codex 36.6% $2.35 13,677
Claude Sonnet 5 medium Claude Code 35.2% $3.43 38,269
Claude Opus 4.7 high Claude Code 35.2% $5.03 30,140
Claude Opus 4.8 low Claude Code 35.1% $2.68 25,194
Claude Opus 4.7 xhigh Claude Code 34.9% $6.87 38,419
GPT-5.5 low Codex 30.6% $1.65 8,855
Claude Opus 4.7 medium Claude Code 28.9% $3.22 21,881
Claude Sonnet 5 low Claude Code 28.7% $2.15 24,817
Claude Opus 4.7 low Claude Code 27.6% $2.47 17,853
GPT-5.4 Mini xhigh Codex 27% $1.52 90,933
Claude Opus 4.6 max Claude Code 26.9% $3.73 28,870
Claude Opus 4.6 high Claude Code 26.6% $3.98 26,677
Claude Sonnet 4.6 max Claude Code 24.3% $2.90 44,192
Claude Opus 4.6 medium Claude Code 23.7% $2.79 19,837
Claude Sonnet 4.6 high Claude Code 23.5% $2.70 41,109
GPT-5.4 Mini high Codex 23.1% $0.95 46,439
Claude Sonnet 4.6 medium Claude Code 21.1% $2.30 33,357
GPT-5.4 Mini medium Codex 20.8% $0.78 34,721
Claude Opus 4.6 low Claude Code 18.3% $2.10 14,888
Claude Sonnet 4.6 low Claude Code 15% $1.65 21,708
GPT-5.4 Mini low Codex 9.25% $0.36 12,066

Number one over time 3 models have held number one with claims included since its release in July 2026

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.

Anthropic Cognition xAI OpenAI
Own leaderboard Replaced by 2.0, Oct 2026
Claude Fable 5 53.5%
Lab claims 4 leaders
GPT-6 Astra 64.5%

Sources: Official FrontierCode 1.1 leaderboard and labs’ model cards

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 Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Astra Claude Fable 5 Claude Opus 4.8 Claude Opus 4.7 GPT-5.4 Mini 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Astra Claude Fable 5

Sources: Official FrontierCode 1.1 leaderboard and labs’ model cards

5 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.

Cognition
xAI
Google DeepMind
OpenAI
Anthropic
DateLaunch postScoreAs the post writes it
Nov 4, 2026 SWE-2 Flash 45.5 FrontierCode 1.1 Main
Oct 7, 2026 Claude Haiku 5.5 46.4 Agentic coding FrontierCode 1.1 (Main) (max)
Sep 28, 2026 Claude Sonnet 5.5 52.1 Agentic coding FrontierCode 1.1 (Main) (Xhigh)
Sep 22, 2026 GPT-6 Luna 42.4 FrontierCode (GPT-6 Luna max)
Sep 22, 2026 GPT-6 Sol 49.3 FrontierCode 1.1 Main (GPT-6 Sol max)
Sep 10, 2026 SWE-2 50 FrontierCode 1.1 Main
Sep 3, 2026 GPT-6 Astra 64.5 FrontierCode 1.1 Extended (score)
Aug 13, 2026 Gemini 3.7 Flash 43.6 FrontierCode 1.1 Main (Score)
Aug 12, 2026 Grok 4.6 61.3 FrontierCode v1.1 (Extended)
Jul 8, 2026 SWE-1.7 42.3 FrontierCode 1.1 Main

7 system cards report it, from 3 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.

DateDocumentScoreAs the document writes itSection
Oct 7, 2026 Claude Haiku 5.5 system card 46.4 FrontierCode 1.1 (Main), max effort (xhigh 45.8, Extended 58.4 at max) (Table 8.1.A) Capabilities
Sep 28, 2026 Claude Sonnet 5.5 system card 46.2 FrontierCode v1.1 (Main) at max effort (xhigh 52.1%, Extended 59.1% max and 64.4% xhigh) (Table 8.1.A, 8.4) Capabilities
Sep 22, 2026 Claude Opus 5.5 system card 54.4 FrontierCode v1.1 (Main) at max effort (best 54.6% at medium. Extended 63.6% at max and 65.3% at medium), mean@5, run by Cognition (Table 8.1.A, 8.4) Capabilities
Sep 1, 2026 Claude Fable 5.1 system card 50.9 FrontierCode 1.1 (Main) overall score at medium effort, Fable 5.1's best effort level; FrontierCode 1.1 Extended 63.6% at medium (8.4) Capabilities
Aug 13, 2026 Gemini 3.7 Flash model card 43.6 FrontierCode 1.1 Main, score (results table, p.5) Capabilities
Aug 12, 2026 Grok 4.6 model card 61.3 FrontierCode v1.1, extended set of 150 samples, extended score, high, run by Cognition (2.3) Capabilities
Jul 24, 2026 Claude Opus 5 system card 53.4 FrontierCode 1.1 (Main) at its best effort level (medium), mean@5, run by Cognition (Extended 63.6% at medium) (Table 8.1.A, 8.4) Capabilities

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.

Anthropic
FrontierCode 2.0 4 leaders
Claude Fable 5.2 50.2%
FrontierCode Replaced by 1.1, Jul 2026
Claude Opus 4.8 34.3%
FrontierCode 1.1 Replaced by 2.0, Oct 2026
Claude Fable 5 53.5%
FrontierCode 2.0 2 leaders
Claude Fable 5.2 50.2%

Source: Official leaderboards

Replaces FrontierCode Replaced by FrontierCode 2.0

More in Coding

All 282 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