Unverified50% confidenceBenchmarkExact time
In DeepSeek's official comparison, when running with the same checkpoint at best effort, its minimal framework scored 72.6 while the open-source code scored 65.5.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/7/2026
First Seen
Valid until: 1/5/2027
Sources
Related Entities
Related Claims
UnverifiedDeepSeek V4.1 claims to achieve 74 on SWE, but the blogger found that this was the best result cherry-picked from multiple tests, with open code testing at 65 and PI testing at 66.78% similarVerifiedOn 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.68% similarUnverifiedThe same DeepSeek v1.1 model scored between 69.8% and 74.2% on Terminal Bench 2.0 under different harnesses.67% similarUnverifiedThe same model's scores on DeepCWE range from 65.5 to 74.2 depending on the agent framework used; the official harness uses the upper end of the range, while under Claude Code's harness the score is only 69.8.66% similarUnverifiedCaveman's fixed benchmark test results from 54 runs show a 33.2% reduction in input tokens, with all 18/18 correctness checks passing.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/984756API
curl https://kongchang.com/api/v1/knowledge/claims/984756MCP
get_claim(id=984756)