Expired50% confidenceFactExact time
On DeepSWE, model scores range from 70% for the top-performing GPT 5.5 to 0% for the bottom-performing Claude Haiku 4.5, with a level of differentiation far greater than the concentrated interval of approximately 30 points on SWE-Bench Pro.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
DeepSWE Benchmark Reveals the Truth: GPT 5.5 Leads Opus 4.7 by a Wide Margin
bilibili7k的每日搬运5/29/2026
Related Claims
UnverifiedOn the DeepSWE benchmark, GPT-5.6 scored 45 at Low, 61 at Medium, 69 at High, 71 at X-High, and 73 at Max across reasoning tiers.67% similarUnverifiedHaiku 4.5 scored 73.3% on SWE Bench, slightly higher than Sonnet 4 and approaching the level of GPT-5 Codex.65% similarUnverified在SWE-bench最初公开结果中GPT-4的解决率仅约1.7%，Devin宣称达到13.86%，SWE-agent达到12.5%64% similarUnverifiedOllama分发的DeepSeek R1各版本均经过GGUF格式量化处理（通常为Q4_K_M或Q8_0精度），将每个参数从16位浮点数压缩至4-8位，显存占用减少50%-75%，推理质量损失通常在5%以内63% similarUnverifiedQwen3-4B is already on par with DeepSeek V3 or GPT-4o on multiple reasoning evaluation metrics.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/47590API
curl https://kongchang.com/api/v1/knowledge/claims/47590MCP
get_claim(id=47590)