- input
- $0.007
- cached
- $0.006
- output
- $6.80
- Context
- 1Mtokens
- Released
- 5mo ago
- Size
- 756Bparameters
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
#24 of 28
- 74.1%
- 73.8%
- 73.6%
- 72.7%
- 69.9%
- 69.6%
- 69.0%
- 68.5%
- 67.5%
- 67.2%
- 43.8%GLM 5.2
Arena Text
Best Arena rating per model · Higher is better
Arena #36
- 1497Gemini 3.8 FlashSimilar price
- 1476GPT-5.2 ChatNearby rank
- 1475GPT-5.4Nearby rank
- 1475GLM 5.2
- 1475
DeepSeek V4.1 FlashNearby rank - 1474GLM 5.3 FlashNearby rank
- 1465GLM 5.1Predecessor
- 1447GPT-5.4 MiniSimilar price
- 1443
InklingSimilar price - 1442Grok 4.3Similar price
+10 vs GLM 5.1 (predecessor) · Ahead of 3 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #37
- 1609GLM 5.3 FlashNearby rank
- 1609
Muse EmberNearby rank - 1608GLM 5.2
- 1594Claude Haiku 5.5Nearby rank
- 1590
Qwen3.8 27BNearby rank - 1583Gemini 3.8 FlashSimilar price
- 1508GLM 5.1Predecessor
- 1414InklingSimilar price
- 1398GPT-5.4 MiniSimilar price
- 1357Grok 4.3Similar price
+100 vs GLM 5.1 (predecessor) · Ahead of 4 of 4 similarly priced models
Arena Search
Best Arena rating per model · Higher is better
Not ranked on Arena Search yet
Agent Arena
Best IPS score per model · Higher is better
Arena #31
- +0.120GPT-6.1 TerraSimilar price
- +0.023Gemini 3.8 FlashSimilar price
- +0.022Muse Spark 1.3Similar price
- +0.005Claude Haiku 5.5Nearby rank
- +0.005
Hy4 previewNearby rank - +0.002GLM 5.2
- +0.001
Qwen3.8 MaxNearby rank - -0.000
MiMo-V2.6-FlashNearby rank - -0.173InklingSimilar price
Ahead of 1 of 4 similarly priced models
ARC-AGI-2
Best score per model · Higher is better
#49 of 85
- 89.2%Gemini 3.8 FlashSimilar price
- 65.1%Grok 4.20Similar price
- 36.5%InklingSimilar price
- 31.1%Gemini 3 Pro PreviewNearby rank
- 22.8%GLM 5.2
- 18.9%GPT-5.4 MiniSimilar price
- 18.3%GPT-5 ProNearby rank
- 4.9%GLM 5Earlier version
+17.9 pts vs GLM 5 (earlier version) · Ahead of 1 of 4 similarly priced models
FrontierMath-Tier-4-v2-Private
Best score per model · Higher is better
#35 of 72
- 31.7%GPT-5.2Nearby rank
- 29.3%Claude Sonnet 5Nearby rank
- 29.3%GLM 5.2
- 29.3%GLM 5.3Nearby rank
- 26.8%Claude Opus 4.6Nearby rank
- 22.0%Gemini 3.8 FlashSimilar price
- 14.6%Grok 4.3Similar price
- 9.8%GPT-5.4 MiniSimilar price
- 4.9%InklingSimilar price
Ahead of 4 of 4 similarly priced models
FrontierMath-Tiers-1-3-v2-Private
Best score per model · Higher is better
#42 of 87
- 68.4%Gemini 3.8 FlashSimilar price
- 62.8%Gemini 3.5 FlashNearby rank
- 59.6%Gemini 3.1 Pro PreviewNearby rank
- 59.2%GLM 5.2
- 58.9%Gemini 3.6 FlashNearby rank
- 58.6%
Mistral Large 4Nearby rank - 51.2%GPT-5.4 MiniSimilar price
- 42.8%Grok 4.3Similar price
- 36.8%GLM 5.1Predecessor
- 33.3%InklingSimilar price
+22.4 pts vs GLM 5.1 (predecessor) · Ahead of 3 of 4 similarly priced models
GPQA diamond
Best score per model · Higher is better
#27 of 161
- 95.4%Gemini 3.8 FlashSimilar price
- 92.6%Gemini 3 Pro PreviewNearby rank
- 92.3%Qwen3.8 Max (0902)Nearby rank
- 91.9%GLM 5.2
- 91.7%DeepSeek V4 Pro 0813Nearby rank
- 91.6%GPT-5.6 LunaNearby rank
- 89.9%GLM 5.1Predecessor
- 88.8%Grok 4.3Similar price
- 88.3%InklingSimilar price
- 86.9%GPT-5.4 MiniSimilar price
+2.0 pts vs GLM 5.1 (predecessor) · Ahead of 3 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#72 of 158
- 98.9%Gemini 3.8 FlashSimilar price
- 93.3%Grok 4.3Similar price
- 93.3%GLM 5.1Predecessor
- 88.9%GPT-5.4 MiniSimilar price
- 88.9%InklingSimilar price
- 86.7%Qwen3.6 35B A3BNearby rank
- 86.7%Qwen3.7 FlashNearby rank
- 86.4%GLM 5.2
- 86.1%Claude Opus 4.5Nearby rank
- 85.8%Claude Sonnet 4.6Nearby rank
-6.9 pts vs GLM 5.1 (predecessor) · Ahead of 0 of 4 similarly priced models
SimpleQA Verified
Best score per model · Higher is better
#59 of 87
- 69.7%Gemini 3.8 FlashSimilar price
- 40.3%InklingSimilar price
- 34.9%Kimi K2.6Nearby rank
- 34.3%Kimi K2.5Nearby rank
- 34.2%GLM 5.2
- 34.0%GLM 5.1Predecessor
- 33.7%Claude Sonnet 5Nearby rank
- 33.2%Grok 4.3Similar price
- 29.4%GPT-5.4 MiniSimilar price
+0.2 pts vs GLM 5.1 (predecessor) · Ahead of 2 of 4 similarly priced models
SWE-bench Verified
Best score per model · Higher is better
#3 of 35
- 83.5%Claude Opus 4.7Nearby rank
- 79.3%Gemini 3.5 FlashNearby rank
- 78.7%GLM 5.2
- 77.6%DeepSeek V4 Pro 0423Similar price
- 77.3%Qwen3.7 MaxSimilar price
- 76.7%Kimi K2.6Similar price
- 75.4%Gemini 3 Flash PreviewSimilar price
- 74.2%GLM 5.1Predecessor
+4.5 pts vs GLM 5.1 (predecessor) · Ahead of 4 of 4 similarly priced models
DeepSWE 1.1
Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22
- pass@4
- 77.0%
- Average cost / attempt
- $3.92
- 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 · max | 43.8% | 77.0% | $3.92 |
| mini-swe-agent · high | 36.3% | 68.1% | $2.84 |
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
FrontierMath-Tiers-1-3-v2-Private
Score by reasoning effort · Higher is better
GPQA diamond
Score by reasoning effort · Higher is better
OTIS Mock AIME 2024-2025
Score by reasoning effort · Higher is better
All published scores
- Arena TextArena #361475
+10 vs GLM 5.1 (predecessor)
- max147595% CI 1471 to 1479 · 54,173 votes
- Arena WebDevArena #371608
+100 vs GLM 5.1 (predecessor)
- max160895% CI 1602 to 1614 · 17,926 votes
- Agent ArenaArena #31+0.002
- max+0.00295% CI -0.006 to +0.010 · 9,341,559 observations
- ARC-AGI-2#49 of 8522.8%
+17.9 pts vs GLM 5 (earlier version)
- unknown22.8%
- FrontierMath-Tier-4-v2-Private#35 of 7229.3%
- max29.3%
Source:Epoch AI evaluation - FrontierMath-Tiers-1-3-v2-Private#42 of 8759.2%
+22.4 pts vs GLM 5.1 (predecessor)
- none42.5%
- low54.7%
- max59.2%
Source:Epoch AI evaluation - GPQA diamond#27 of 16191.9%
+2.0 pts vs GLM 5.1 (predecessor)
- none71.2%
- low87.9%
- max91.9%
Source:Epoch AI evaluation - OTIS Mock AIME 2024-2025#72 of 15886.4%
-6.9 pts vs GLM 5.1 (predecessor)
- none28.9%
- low75.6%
- max86.4%
Source:Epoch AI evaluation - SimpleQA Verified#59 of 8734.2%
+0.2 pts vs GLM 5.1 (predecessor)
- max34.2%
Source:Epoch AI evaluation - SWE-bench Verified#3 of 3578.7%
+4.5 pts vs GLM 5.1 (predecessor)
- max78.7%
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.