Benchmark Almanac

GameASG-Bench

GPT-6 Astra 55.3% leads since September 2026
Not in model cards

Agents build 47 browser games from written requirements. Static checks inspect each HTML file, and browser tests check controls, gameplay and state changes.

The September 18, 2026 paper compares nine model and harness combinations. GPT-6-Astra with Codex CLI led with 26 of 47 games passing (55.3%).

Made by
AReaL project
Released
September 2026
Kind
Building whole projects
Level
University
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code, Actions
Interaction
Agentic
Environment
Terminal, Browser
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
9
Entries in its paper, 8 models we track

GPT-6 Astra leads with 55.3% among models released by September 2026 GPT-6 Astra leads with 55.3% among models released by September 2026

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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. This view shows only models we track, each at its best.

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:

  1. 1 GPT-6 Astra Maximum reasoning effort 55.3%
  2. 2 Claude Opus 5 Maximum reasoning effort 51.1%
  3. 3 GPT-5.6 Sol Maximum reasoning effort 44.7%
  4. 4 DeepSeek-V4-Flash High reasoning effort 40.4%
  5. 5 DeepSeek-V4-Pro Maximum reasoning effort 31.9%
  6. 6 Kimi K3 Maximum reasoning effort 31.9%
  7. 7 GLM-5.2 Maximum reasoning effort 23.4%
  8. 8 Hunyuan-3 Maximum reasoning effort 21.3%
  9. 9 MiniMax-M3 Maximum reasoning effort 14.9%

Source: The GameASG-Bench paper

#EntryScoreSettingDateSource
1 GPT-6 Astra 55.3% Maximum reasoning effort, Codex CLI Sep 2026 Official leaderboard
2 Claude Opus 5 51.1% Maximum reasoning effort, Claude Code Jul 2026 Official leaderboard
3 GPT-5.6 Sol 44.7% Maximum reasoning effort, Codex CLI Jul 2026 Official leaderboard
4 DeepSeek-V4-Flash 40.4% High reasoning effort, Claude Code Apr 2026 Official leaderboard
5 DeepSeek-V4-Pro 31.9% Maximum reasoning effort, Claude Code Apr 2026 Official leaderboard
6 Kimi K3 31.9% Maximum reasoning effort, Claude Code Jul 2026 Official leaderboard
7 GLM-5.2 23.4% Maximum reasoning effort, Claude Code Jun 2026 Official leaderboard
8 Hunyuan-3 21.3% Maximum reasoning effort, Claude Code Official leaderboard
9 MiniMax-M3 14.9% Maximum reasoning effort, Claude Code Jun 2026 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-6 Astra 55.3% Maximum reasoning effort, Codex CLI Sep 2026 Official leaderboard
2 Claude Opus 5 51.1% Maximum reasoning effort, Claude Code Jul 2026 Official leaderboard
3 GPT-5.6 Sol 44.7% Maximum reasoning effort, Codex CLI Jul 2026 Official leaderboard
4 DeepSeek-V4-Flash 40.4% High reasoning effort, Claude Code Apr 2026 Official leaderboard
5 DeepSeek-V4-Pro 31.9% Maximum reasoning effort, Claude Code Apr 2026 Official leaderboard
6 Kimi K3 31.9% Maximum reasoning effort, Claude Code Jul 2026 Official leaderboard
7 GLM-5.2 23.4% Maximum reasoning effort, Claude Code Jun 2026 Official leaderboard
8 MiniMax-M3 14.9% Maximum reasoning effort, Claude Code Jun 2026 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s paper, each ranking models on one part of the test, such as L1, L2 and L2 P0. DeepSeek-V4-Flash leads 4 of them, and GPT-6 Astra leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

L1

8 models we track have a score here.

  1. 1 DeepSeek-V4-Flash ↑3 100%
  2. 2 DeepSeek-V4-Pro ↑3 99.6%
  3. 3 Claude Opus 5 ↓1 99.6%
  4. 4 GPT-5.6 Sol ↓1 98.9%
  5. 5 GLM-5.2 ↑2 98.8%
  6. 6 MiniMax-M3 ↑2 98.3%
  7. 7 GPT-6 Astra ↓6 98%
  8. 8 Kimi K3 ↓2 97.7%

The best score, 55.3%, costs $4.44 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 OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6 Astra, Maximum reasoning effort GPT-5.6 Sol, Maximum reasoning effort DeepSeek-V4-Flash, Maximum reasoning effort 0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6 Astra, Maximum reasoning effort GPT-5.6 Sol, Maximum reasoning effort

Source: The GameASG-Bench paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra Maximum reasoning effort Codex CLI 55.3% $4.44 55,500
Claude Opus 5 Maximum reasoning effort Claude Code 51.1% $12 142,400
GPT-5.6 Sol Maximum reasoning effort Codex CLI 44.7% $1.79 45,700
DeepSeek-V4-Flash High reasoning effort Claude Code 40.4%
DeepSeek-V4-Flash Maximum reasoning effort Claude Code 38.3% $0.25 203,600
DeepSeek-V4-Flash Maximum reasoning effort Codex CLI 38.3%
DeepSeek-V4-Pro Maximum reasoning effort Claude Code 31.9% $0.74 202,900
Kimi K3 Maximum reasoning effort Claude Code 31.9% $2.55 72,300
DeepSeek-V4-Flash 60 turns Claude Code 27.7%
GLM-5.2 Maximum reasoning effort Claude Code 23.4% $0.62 47,000
DeepSeek-V4-Flash Files + syntax checking Claude Code 19.1%
DeepSeek-V4-Flash No tools (serialized API) Serialized API 14.9%
DeepSeek-V4-Flash 30 turns Claude Code 14.9%
DeepSeek-V4-Flash Low reasoning effort Claude Code 14.9%
MiniMax-M3 Maximum reasoning effort Claude Code 14.9% $0.41 71,100
DeepSeek-V4-Flash File read/write only Claude Code 12.8%

The best score, 55.3%, costs $4.44 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 OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6 Astra, Maximum reasoning effort GPT-5.6 Sol, Maximum reasoning effort DeepSeek-V4-Flash, Maximum reasoning effort 0% 20% 40% 60% $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6 Astra, Maximum reasoning effort GPT-5.6 Sol, Maximum reasoning effort

Source: The GameASG-Bench paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra Maximum reasoning effort Codex CLI 55.3% $4.44 55,500
Claude Opus 5 Maximum reasoning effort Claude Code 51.1% $12 142,400
GPT-5.6 Sol Maximum reasoning effort Codex CLI 44.7% $1.79 45,700
DeepSeek-V4-Flash High reasoning effort Claude Code 40.4%
DeepSeek-V4-Flash Maximum reasoning effort Claude Code 38.3% $0.25 203,600
DeepSeek-V4-Flash Maximum reasoning effort Codex CLI 38.3%
DeepSeek-V4-Pro Maximum reasoning effort Claude Code 31.9% $0.74 202,900
Kimi K3 Maximum reasoning effort Claude Code 31.9% $2.55 72,300
DeepSeek-V4-Flash 60 turns Claude Code 27.7%
GLM-5.2 Maximum reasoning effort Claude Code 23.4% $0.62 47,000
DeepSeek-V4-Flash Files + syntax checking Claude Code 19.1%
DeepSeek-V4-Flash No tools (serialized API) Serialized API 14.9%
DeepSeek-V4-Flash 30 turns Claude Code 14.9%
DeepSeek-V4-Flash Low reasoning effort Claude Code 14.9%
MiniMax-M3 Maximum reasoning effort Claude Code 14.9% $0.41 71,100
DeepSeek-V4-Flash File read/write only Claude Code 12.8%

2 models have held number one in the benchmark’s paper since its release in September 2026

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

Anthropic OpenAI
Benchmark paper Not tested since Sep 2026
GPT-6 Astra 55.3%

Source: The GameASG-Bench paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) GPT-6 Astra Claude Opus 5 GPT-5.6 Sol DeepSeek-V4-Flash 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) GPT-6 Astra GPT-5.6 Sol

Source: Official GameASG-Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Coding

All 523 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

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (9 entries).
  • Pass Models scored 8 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from September 2026, for GPT-6 Astra.
  • Note Frontier models Only 1 of the top 10 has a score. The strongest tested is GPT-6 Astra (#4).
  • Pass Headroom The best is 55.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 16 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check