- Context
- 1.1Mtokens
- Released
- 2mo ago
Cost/1M tokens
Benchmarks
DeepSWE is the primary coding benchmark where an exact model evaluation exists. Arena leaderboards and other sourced evaluations appear below; missing scores are not zero.
How it ranks
Against its predecessor, similarly priced models and the models ranked around it.
DeepSWE 1.1
Best pass@1 per model · Higher is better
#1 of 28
- 74.1%
GPT-6 Astra - 73.8%
- 73.6%
- 72.7%
- 69.9%
- 69.6%
- 69.0%
- 68.5%
- 67.5%
- 67.2%
Arena Text
Best Arena rating per model · Higher is better
Arena #41
- 1504Claude Fable 5Similar price
- 1503Claude Opus 4.6Similar price
- 1503Claude Fable 5.1Similar price
- 1485GPT-5.6 SolPredecessor
- 1475
DeepSeek V4.1 FlashNearby rank - 1474GLM 5.3 FlashNearby rank
- 1474GPT-6 Astra
- 1474Claude Opus 4.5Nearby rank
- 1473GPT-5.5 InstantNearby rank
- 1450Claude Opus 4.1Similar price
-11 vs GPT-5.6 Sol (predecessor) · Ahead of 1 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #5
- 1836Claude Fable 5.2Similar price
- 1773Claude Sonnet 5.5Nearby rank
- 1768Gemini 4 CarbonNearby rank
- 1764GPT-6 Astra
- 1743GPT-6.1 SolNearby rank
- 1742Claude Fable 5.1Similar price
- 1625Claude Fable 5Similar price
- 1615GPT-5.6 SolPredecessor
- 1389Claude Opus 4.1Similar price
+149 vs GPT-5.6 Sol (predecessor) · Ahead of 3 of 4 similarly priced models
Arena Vision
Best Arena rating per model · Higher is better
Arena #21
- 1309Claude Fable 5Similar price
- 1300Claude Opus 4.7Similar price
- 1299Claude Opus 4.6Similar price
- 1289Claude Fable 5.1Similar price
- 1288GPT-5.6 SolPredecessor
- 1285Claude Opus 5Nearby rank
- 1285GPT-5.5Nearby rank
- 1283GPT-6 Astra
- 1282GPT-5.4Nearby rank
- 1281Gemini 3.5 FlashNearby rank
Level with GPT-5.6 Sol (predecessor, overlapping 95% CIs) · Ahead of 0 of 4 similarly priced models
Arena Search
Best Arena rating per model · Higher is better
Arena #3
- 1255GPT-5.6 SolPredecessor
- 1248GPT-6.1 SolNearby rank
- 1245GPT-6 Astra
- 1229GPT-6 SolNearby rank
- 1229Claude Fable 5Similar price
- 1227Claude Fable 5.1Similar price
- 1213Claude Opus 5Similar price
- 1210Claude Opus 4.7Similar price
Level with GPT-5.6 Sol (predecessor, overlapping 95% CIs) · Ahead of 4 of 4 similarly priced models
Arena Document
Best Arena rating per model · Higher is better
Arena #17
- 1500Claude Fable 5Similar price
- 1498Claude Fable 5.1Similar price
- 1497Claude Opus 4.6Similar price
- 1494Claude Opus 4.7Similar price
- 1486GPT-5.6 SolPredecessor
- 1474GPT-5.6 TerraNearby rank
- 1474
Muse Spark 1.3Nearby rank - 1474GPT-6 Astra
- 1471GPT-5.4Nearby rank
- 1469Gemini 3.8 FlashNearby rank
Level with GPT-5.6 Sol (predecessor, overlapping 95% CIs) · Ahead of 0 of 4 similarly priced models
Agent Arena
Best IPS score per model · Higher is better
Arena #6
- +0.304Claude Fable 5.2Similar price
- +0.160Claude Sonnet 5.5Nearby rank
- +0.148Claude Fable 5.1Similar price
- +0.145GPT-6 Astra
- +0.125GPT-6.1 SolNearby rank
- +0.120GPT-6.1 TerraNearby rank
- +0.073Claude Fable 5Similar price
- +0.072Claude Opus 5Similar price
- +0.046GPT-5.6 SolPredecessor
+0.099 vs GPT-5.6 Sol (predecessor) · Ahead of 2 of 4 similarly priced models
ARC-AGI-2
Best score per model · Higher is better
#1 of 85
- 95.0%GPT-6 Astra
- 94.2%Gemini 4 CarbonNearby rank
- 94.2%GPT-6.1 SolNearby rank
- 93.8%Claude Fable 5.2Similar price
- 92.5%GPT-5.6 SolPredecessor
- 90.0%Claude Fable 5.1Similar price
- 89.2%Claude Fable 5Similar price
- 88.3%Claude Opus 5Similar price
+2.5 pts vs GPT-5.6 Sol (predecessor) · Ahead of 4 of 4 similarly priced models
FrontierMath-Tier-4-v2-Private
Best score per model · Higher is better
#3 of 72
- 100.0%Claude Fable 5.2Similar price
- 100.0%GPT-6.1 SolNearby rank
- 97.6%GPT-6 Astra
- 95.1%Gemini 4 CarbonNearby rank
- 95.0%Claude Opus 5.5Nearby rank
- 90.2%Claude Fable 5Similar price
- 87.8%Claude Fable 5.1Similar price
- 82.9%GPT-5.6 SolPredecessor
- 2.4%Claude Opus 4.1Similar price
+14.7 pts vs GPT-5.6 Sol (predecessor) · Ahead of 3 of 4 similarly priced models
FrontierMath-Tiers-1-3-v2-Private
Best score per model · Higher is better
#2 of 87
- 95.1%Claude Fable 5.2Similar price
- 93.7%GPT-6 Astra
- 93.7%GPT-6.1 SolNearby rank
- 91.2%Claude Opus 5.5Nearby rank
- 90.2%Claude Fable 5.1Similar price
- 89.1%GPT-5.6 SolPredecessor
- 87.0%Claude Fable 5Similar price
- 12.6%Claude Opus 4.1Similar price
+4.6 pts vs GPT-5.6 Sol (predecessor) · Ahead of 3 of 4 similarly priced models
GPQA diamond
Best score per model · Higher is better
#3 of 161
- 96.5%Gemini 4 CarbonNearby rank
- 96.4%Gemini 4 ArgonNearby rank
- 95.8%GPT-6 Astra
- 95.6%Claude Sonnet 5.5Nearby rank
- 95.4%Gemini 3.8 FlashNearby rank
- 93.9%Claude Fable 5.2Similar price
- 93.9%Claude Opus 5Similar price
- 93.5%GPT-5.6 SolPredecessor
- 85.9%Claude Fable 5Similar price
- 77.3%Claude Opus 4.1Similar price
+2.3 pts vs GPT-5.6 Sol (predecessor) · Ahead of 4 of 4 similarly priced models
Humanity's Last Exam
Best score per model · Higher is better
#1 of 34
- 54.8%GPT-6 Astra
- 46.5%Claude Fable 5.1Similar price
- 46.4%Gemini 3.1 Pro PreviewNearby rank
- 36.2%GPT-5.4Earlier version
- 36.2%Claude Opus 4.7Similar price
- 34.4%Claude Opus 4.6Similar price
- 7.9%Claude Opus 4.1Similar price
+18.6 pts vs GPT-5.4 (earlier version) · Ahead of 4 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#1 of 158
- 100.0%Claude Fable 5Similar price
- 100.0%Claude Fable 5.1Similar price
- 100.0%Gemini 4 CarbonNearby rank
- 100.0%GPT-5.6 SolPredecessor
- 100.0%GPT-6 Astra
- 100.0%GPT-6 SolNearby rank
- 100.0%GPT-6.1 SolNearby rank
- 99.7%Claude Fable 5.2Similar price
- 68.9%Claude Opus 4.1Similar price
Level with GPT-5.6 Sol (predecessor) · Ahead of 2 of 4 similarly priced models
SimpleQA Verified
Best score per model · Higher is better
#4 of 87
- 94.0%Claude Fable 5.2Similar price
- 77.7%Gemini 4 CarbonNearby rank
- 76.0%Gemini 4 ArgonNearby rank
- 75.6%GPT-6 Astra
- 73.9%GPT-6.1 SolNearby rank
- 73.5%Gemini 3.1 Pro PreviewNearby rank
- 70.8%Claude Fable 5.1Similar price
- 70.7%Claude Fable 5Similar price
- 69.7%GPT-5.6 SolPredecessor
- 59.9%Claude Opus 5Similar price
+5.9 pts vs GPT-5.6 Sol (predecessor) · Ahead of 3 of 4 similarly priced models
DeepSWE 1.1
Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22
- pass@4
- 80.5%
- Average cost / attempt
- $4.43
- Benchmark runs
- 4
Score vs cost by reasoning effort
Average cost per attempt (USD) against pass rate
All published configurations
| Harness / reasoning | pass@1 | pass@4 | Cost / attempt |
|---|---|---|---|
| mini-swe-agent · xhigh | 74.1% | 80.5% | $4.43 |
| mini-swe-agent · high | 73.2% | 82.3% | $3.92 |
| mini-swe-agent · max | 73.2% | 79.6% | $7.50 |
| mini-swe-agent · medium | 72.8% | 82.3% | $3.08 |
| mini-swe-agent · low | 67.0% | 79.6% | $1.60 |
pass@1 is the success rate across scored attempts; pass@4 is the share of tasks solved in at least one of four attempts. Costs are reported by DeepSWE for each configuration. These results are not directly comparable to evaluations using different harnesses. Official source JSON.
Reasoning effort
ARC-AGI-2
Score by reasoning effort · Higher is better
FrontierMath-Tier-4-v2-Private
Score by reasoning effort · Higher is better
All published scores
- Arena TextArena #411474
-11 vs GPT-5.6 Sol (predecessor)
- max147495% CI 1469 to 1480 · 18,315 votes
- Arena WebDevArena #51764
+149 vs GPT-5.6 Sol (predecessor)
- max176495% CI 1756 to 1773 · 8,425 votes
- Arena VisionArena #211283
Level with GPT-5.6 Sol (predecessor, overlapping 95% CIs)
- max128395% CI 1274 to 1293 · 5,517 votes
- Arena SearchArena #31245
Level with GPT-5.6 Sol (predecessor, overlapping 95% CIs)
- max124595% CI 1237 to 1253 · 17,456 votes
- Arena DocumentArena #171474
Level with GPT-5.6 Sol (predecessor, overlapping 95% CIs)
- max147495% CI 1465 to 1483 · 5,723 votes
- Agent ArenaArena #6+0.145
+0.099 vs GPT-5.6 Sol (predecessor)
- max+0.14595% CI +0.125 to +0.166 · 1,353,961 observations
- ARC-AGI-2#1 of 8595.0%
+2.5 pts vs GPT-5.6 Sol (predecessor)
- none59.6%
- low85.4%
- medium92.1%
- high92.1%
- xhigh93.3%
- max95.0%
- FrontierMath-Tier-4-v2-Private#3 of 7297.6%
+14.7 pts vs GPT-5.6 Sol (predecessor)
- none82.9%
- low87.8%
- medium97.6%
- high97.6%
- xhigh97.6%
- max97.6%
Source:Epoch AI evaluation - FrontierMath-Tiers-1-3-v2-Private#2 of 8793.7%
+4.6 pts vs GPT-5.6 Sol (predecessor)
- max93.7%
Source:Epoch AI evaluation - GPQA diamond#3 of 16195.8%
+2.3 pts vs GPT-5.6 Sol (predecessor)
- max95.8%
Source:Epoch AI evaluation - Humanity's Last Exam#1 of 3454.8%
+18.6 pts vs GPT-5.4 (earlier version)
- unknown54.8%
- OTIS Mock AIME 2024-2025#1 of 158100.0%
Level with GPT-5.6 Sol (predecessor)
- max100.0%
Source:Epoch AI evaluation - SimpleQA Verified#4 of 8775.6%
+5.9 pts vs GPT-5.6 Sol (predecessor)
- max75.6%
Source:Epoch AI evaluation
Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.
Arena ratings: Arena (lmarena-ai/leaderboard-dataset), licensed CC BY 4.0. Ratings come from pairwise human votes and are only comparable within one leaderboard; the interval is Arena's 95% confidence interval. Model names are matched to the Vertical Tabs catalog.