korshunov.ai Everything that matters in ML
EN
Benchmark · agentic

DeepSWE

25 results 20 models

DataCurve's long-horizon real-repo coding-agent suite (headline metric = pass@1 resolve rate); Artificial Analysis swapped SWE-bench Pro out of its Coding Agent Index for it. Not a SWE-bench variant.

0 23.5 47 70.5 94 2026-07-27 2026-08-18 2026-09-10 Kimi K3 · 68.5 · 2026-07-27 GPT-5.6 Sol · 72.7 · 2026-07-27 GPT-5.6 Sol · 85.8 · 2026-08-18 Gemini 3.6 Flash · 49 · 2026-07-27 DeepSeek-V4-Flash · 54.4 · 2026-07-31 DeepSeek-V4-Flash-0731 · 54.4 · 2026-07-31 Qwen3.8-Max · 56.6 · 2026-08-03 DeepSeek-V4 Flash 0731 · 53.3 · 2026-08-07 GPT-5.6 Luna · 67.2 · 2026-08-07 Grok 4.6 · 65.9 · 2026-08-13 DeepSeek-V4-Pro · 62.7 · 2026-08-13 Gemini 3.7 Flash · 65.3 · 2026-08-13 GLM-5.3 · 66.9 · 2026-08-17 GLM-5.3 · 69 · 2026-08-22 DeepSeek V4 Pro 0813 · 88.5 · 2026-08-18 DeepSeek V4 Pro 0813 · 62.8 · 2026-08-18 Claude Fable 5 · 69.7 · 2026-08-18 Claude Fable 5 · 69.7 · 2026-08-22 Ornith-1.5 · 56 · 2026-08-19 Ornith-1.5 · 56 · 2026-08-20 DeepSeek-V4-Flash-Vision-Exp · 59.3 · 2026-08-21 GLM-5.3-Flash · 63 · 2026-08-26 Muse Spark 1.3 · 75.4 · 2026-09-04 GPT-6 Astra · 74.1 · 2026-09-04 DeepSeek-V4.1-Flash · 74.2 · 2026-09-10
Kimi K3 GPT-5.6 Sol Gemini 3.6 Flash DeepSeek-V4-Flash DeepSeek-V4-Flash-0731 Qwen3.8-Max DeepSeek-V4 Flash 0731 GPT-5.6 Luna Grok 4.6 DeepSeek-V4-Pro Gemini 3.7 Flash GLM-5.3 DeepSeek V4 Pro 0813 Claude Fable 5 Ornith-1.5 DeepSeek-V4-Flash-Vision-Exp GLM-5.3-Flash Muse Spark 1.3 GPT-6 Astra DeepSeek-V4.1-Flash
Timeline
Date Model Score Source
2026-09-10 DeepSeek-V4.1-Flash 74.2% DeepSeek releases V4.1-Flash model with native multimodal support
2026-09-04 GPT-6 Astra 74.1% OpenAI releases GPT-6 Astra, a computer-use model with 1.05M context
2026-09-04 Muse Spark 1.3 75.4% Meta releases Muse Spark 1.3 with fewer tool calls and tokens
2026-08-26 GLM-5.3-Flash 63.0% Ox Alpha is GLM-5.3-Flash
2026-08-22 GLM-5.3 69.0% GLM-5.3 matches Claude Fable 5 accuracy on DeepSWE at a fifth of the cost
2026-08-22 Claude Fable 5 69.7% GLM-5.3 matches Claude Fable 5 accuracy on DeepSWE at a fifth of the cost
2026-08-21 DeepSeek-V4-Flash-Vision-Exp 59.3% DeepSeek releases DeepSeek-V4-Flash-Vision-Exp multimodal model
2026-08-20 Ornith-1.5 56.0% Z.ai proposes post-training scaling law and releases GLM 5.3; Ornith-1.5 launches with self-improvement
2026-08-19 Ornith-1.5 56.0% Ornith-1.5 releases 9B, 35B-A3B, and 397B models with self-improving training
2026-08-18 DeepSeek V4 Pro 0813 62.8% DeepSeek V4 Pro 0813 vs Claude Fable 5 on DeepSWE: Cost, Coding, and Routing
2026-08-18 Claude Fable 5 69.7% DeepSeek V4 Pro 0813 vs Claude Fable 5 on DeepSWE: Cost, Coding, and Routing
2026-08-18 GPT-5.6 Sol 85.8% DeepSeek V4 Pro and GPT-5.6 Sol cascade solves 83% of DeepSWE tasks at $3.35
2026-08-18 DeepSeek V4 Pro 0813 88.5% DeepSeek V4 Pro and GPT-5.6 Sol cascade solves 83% of DeepSWE tasks at $3.35
2026-08-17 GLM-5.3 66.9% Z.ai launches GLM-5.3 coding model; Cursor joins SpaceXAI
2026-08-13 Gemini 3.7 Flash 65.3% Google introduces Gemini 3.7 Flash with improved coding and lower cost
2026-08-13 DeepSeek-V4-Pro 62.7% DeepSeek-V4-Pro GA release enhances agent capabilities and adds Responses API support
2026-08-13 Grok 4.6 65.9% SpaceXAI releases Grok 4.6 with 500K context and improved agentic reasoning
2026-08-07 DeepSeek-V4 Flash 0731 53.3% DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Coding
2026-08-07 GPT-5.6 Luna 67.2% DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Coding
2026-08-03 Qwen3.8-Max 56.6% Alibaba releases Qwen3.8-Max, a 2.4T-parameter MoE model
2026-07-31 DeepSeek-V4-Flash-0731 54.4% DeepSeek releases V4-Flash-0731 with major agentic gains at lower cost
2026-07-31 DeepSeek-V4-Flash 54.4% DeepSeek releases DeepSeek-V4-Flash in public beta with enhanced agent capabilities
2026-07-27 Gemini 3.6 Flash 49.0% Google releases Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
2026-07-27 Kimi K3 68.5% Kimi K3 beats GPT-5.6 Sol on DeepSWE pass@4 and costs 64% less
2026-07-27 GPT-5.6 Sol 72.7% Kimi K3 beats GPT-5.6 Sol on DeepSWE pass@4 and costs 64% less