Home Model series DeepSeek V3 DeepSeek V3 · SWE-bench Verified · Agent mode; task snapshot unspecified · Release history DeepSeek V3 · SWE-bench Verified · Agent mode; task snapshot unspecified · Release history View scores, configurations, and disclosures for SWE-bench Verified · Agent mode; task snapshot unspecified across 3 DeepSeek V3 releases, ordered by release date.
Differences are calculated only for unique, active, exact values between direct predecessors when the unit, metric, and explicitly recorded configuration and conditions match.
2024-12-26 · 1 score record for this release, including 1 exact value (may include historical records)
DeepSeek-V3 · SWE-bench Verified scores and configurations Metric and score Configuration and evaluation conditions Sources Resolved 42% Reported 42 · Unit % Published Unconfirmed
DeepSeek-V3 Chat Settings Chat checkpoint; source table limits output to 8K and repeats benchmarks under the card protocol. Unknowns Exact temperature for each row is not stated in the table. Tools Tool access not recorded Reasoning budget Output limit 8192 tokens Framework not recorded Protocol not specified Conditions Chat-model standard benchmarks; source note states output length limited to 8K. View this score in rankings
2025-08-21 · 1 score record for this release, including 1 exact value (may include historical records)
DeepSeek-V3.1 · SWE-bench Verified scores and configurations Metric and score Configuration and evaluation conditions Sources Resolved 66% Reported 66 · Unit % Published Unconfirmed
DeepSeek-V3.1-NonThinking Settings Non-thinking mode as named in the official table. Unknowns none recorded Tools Tool access not recorded Reasoning budget not recorded Framework not recorded Protocol not specified Conditions SWE Verified (Agent mode) View this score in rankings
2025-09-22 · 1 score record for this release, including 1 exact value (may include historical records)
Records for the direct predecessor do not provide matching conditions for a valid difference.
DeepSeek-V3.1-Terminus · SWE-bench Verified scores and configurations Metric and score Configuration and evaluation conditions Sources Resolved 68.4% Reported 68.4 · Unit % Published Unconfirmed
DeepSeek-V3.1-Terminus reported comparison Settings The table does not specify an effort setting; search-agent template/tool-set changed as stated in the card. Unknowns Do not infer a reasoning mode from the predecessor comparison. Tools Tool access not recorded Reasoning budget not recorded Framework not recorded Protocol not specified Conditions Metric/harness not restated in this comparison table. View this score in rankings