deepseek-v4.1POST
DeepSeek V4.1 Flash official benchmarks: 77.0 on DeepSWE v1.1, above Opus 5
DeepSeek says V4.1-Flash-1017 beats Claude Opus 5 on agentic coding: 77.0 vs 74.0 on DeepSWE v1.1 and 35.4 vs 28.6 on Agents' Last Exam, at about $0.20 per task versus $0.50 before.
Qualiteg Inc.
@Qualiteg_globalArchived 13 days agoOriginally posted 2026.10.18
Qualiteg Inc. @Qualiteg_globalOriginally posted 2026.10.18
DeepSeek says its V4.1-Flash-1017 beats Claude Opus 5 on agentic coding. - DeepSWE v1.1 hits 77.0, vs Opus 5's 74.0 - Agents' Last Exam reaches 35.4, vs Opus 5's 28.6 - Costs about $0.20 per task, a third of the prior version's $0.50 Link in reply. #AI #OpenModels
x.com/qualiteg_global/status/2111601615367217651Archived 2026.10.18 16:33