DeepSeek-V3.1-Terminus: benchmark scores, price and ranking
DeepSeek-V3.1-Terminus from DeepSeek ranks #440 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.4), based on 5 benchmark results captured up to 2026-10-13.
- Organisation
- DeepSeek
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2025-09-22 (source: epoch)
- Cheapest API price
- $0.270 in / $1.00 out per million tokens (via deepinfra, observed 2026-10-13T20:47Z)
- Aggregate rank
- #440 of 743 (score 47.4)
Relative to other models, DeepSeek-V3.1-Terminus is strongest on DTBench (#48 of 146) and weakest on SciCode (#61 of 106).
Across the 5 benchmarks where at least ten models are scored, DeepSeek-V3.1-Terminus is in the top 10% on 0 and in the top half on 1.
Its cheapest listed input price, $0.270 per million tokens, is lower than that of 62% of the 427 models with a listed price.
DeepSeek-V3.1-Terminus benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| DTBench | 81.3% | #48 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| ALE-Bench | 745 | #41 of 79 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CritPt | 1.7% | #61 of 115 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 28.6% | #58 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SciCode | 40.6% | #61 of 106 | Epoch AI Benchmarking Hub | 2026-10-13 |
About DeepSeek-V3.1-Terminus
DeepSeek-V3.1-Terminus is an open-weight model that can run on your own hardware from DeepSeek. It was released on 2025-09-22, according to Epoch AI. The cheapest API listing we track charges $0.270 per million input tokens and $1.00 per million output tokens, through deepinfra. We hold 5 benchmark results for it across 4 areas, and none is in the top tenth of its board.
Where DeepSeek-V3.1-Terminus stands
On our aggregate score, DeepSeek-V3.1-Terminus is #440 of 743 ranked models, around the middle of the table. Its score is 47.4; the scale is centred on 50, so it sits 2.6 points below the typical model. That beats 41% of the ranked field. Among the 33 ranked DeepSeek models it is #24. The next one up is Deepseek R1 (05/28) at 47.5. The next one down is DeepSeek-R1-Distill-Qwen-14B at 47.0.
DeepSeek-V3.1-Terminus, benchmark by benchmark
DTBench puts DeepSeek-V3.1-Terminus at 81.3%, #48 of 146. That beats 67% of the models tested, but 17.6 points short of Claude Opus 5.5 (max) (98.9%). The results either side are Claude Opus 4 (81.6%) and qwen-plus-2025-04-28 (81.1%). It covers dynamic tool-use benchmark. The board median is 68.8%, so DeepSeek-V3.1-Terminus is 12.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
DeepSeek-V3.1-Terminus lands below the middle on ALE-Bench: 745, #41 of 79, with 51% of the field scoring higher. GPT-6 Astra (max) leads at 2,951; the closest result above is Mistral Medium 3.5 at 764. It covers AtCoder heuristic-optimisation contests (performance rating). The board median is 764, so DeepSeek-V3.1-Terminus is 19 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
DeepSeek-V3.1-Terminus lands below the middle on CritPt: 1.7%, #61 of 115, with 52% of the field scoring higher. GPT-5.6 Sol (max) leads at 32.3%; the closest result above is Step 3.7 Flash at 2.3%. It covers unpublished research-level physics problems. The board median is 2.9%, so DeepSeek-V3.1-Terminus is 1.1 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
DeepSeek-V3.1-Terminus lands below the middle on LMCA: 28.6%, #58 of 108, with 53% of the field scoring higher. Claude Opus 5.5 (max) leads at 68.2%; the closest result above is DeepSeek-V3.2 (Thinking; Novita) at 28.8%. It covers language-model capability assessment. The board median is 29.2%, so DeepSeek-V3.1-Terminus is 0.6 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
DeepSeek-V3.1-Terminus lands below the middle on SciCode: 40.6%, #61 of 106, with 57% of the field scoring higher. Claude Opus 5.5 (max) leads at 66.9%; the closest result above is Qwen 3.6 Plus (2026-04-02) at 40.7%. It covers scientific research code. The board median is 44.3%, so DeepSeek-V3.1-Terminus is 3.7 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Software engineering: uneven, ahead of 45% of the field on average across 2 benchmarks. Its best board here is ALE-Bench (#41 of 79) and its weakest is SciCode (#61 of 106).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so DeepSeek-V3.1-Terminus costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on deepseek-chat ($0.280) and Qwen3.7 Plus ($0.282).
Release history
The DeepSeek model released before it was DeepSeek-V3.1 on 2025-08-21, which scores 48.0. The next one was deepseek-chat on 2025-12-01, at 46.5.
Reading these numbers
The 5 scores on this page come from 1 source: Epoch AI Benchmarking Hub (5). No official leaderboard supplied any of them, so treat them as third-party measurements. With 5 results the aggregate rank can move a lot when one new score arrives.
Models ranked nearby
- #437 Magistral Small (47.4)
- #438 mini-SWE-agent + Gemini 3.7 Flash (47.4)
- #439 Mercury 2.5 (high) (47.4)
- #441 gemini-1.5-flash-latest (47.4)
- #442 claude-haiku-4-5-20251001_32K (47.3)
- #443 Gemma 4 31B IT (47.3)
More from DeepSeek
- DeepSeek V4 Pro 0813 (max) (#23, 54.3)
- DeepSeek V4 Flash 0731 (max) (#42, 53.1)
- DeepSeek-V3.2-Exp (high) (#67, 52.4)
- deepseek-v4.1-flash-max (#75, 52.2)
- DeepSeek-V2.5-1210 (#101, 51.7)
- DeepSeek v4 Pro (max) (#108, 51.6)
Compare it on the live leaderboard → · How the aggregate score works