DeepSeek-V3 (Mar 2025): benchmark scores, price and ranking

DeepSeek-V3 (Mar 2025) from DeepSeek ranks #647 of 743 LLMs on the LLMs Tiger aggregate benchmark score (44.1), based on 11 benchmark results captured up to 2026-10-13.

Organisation: DeepSeek
Type: Open-weight / local model
Open weights: Yes
Released: 2025-03-24 (source: epoch)
Cheapest API price: $0.200 in / $0.200 out per million tokens (via hyperbolic, observed 2026-10-13T19:47Z)
Aggregate rank: #647 of 743 (score 44.1)

Relative to other models, DeepSeek-V3 (Mar 2025) is strongest on Epoch ECI (#68 of 204) and weakest on LMCA (#86 of 108).

Across the 11 benchmarks where at least ten models are scored, DeepSeek-V3 (Mar 2025) is in the top 10% on 0 and in the top half on 1.

Its cheapest listed input price, $0.200 per million tokens, is lower than that of 67% of the 427 models with a listed price.

DeepSeek-V3 (Mar 2025) benchmark results

DeepSeek-V3 (Mar 2025) benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
Epoch ECI	136	#68 of 204	Epoch AI Benchmarking Hub	2026-10-13
METR time horizon	23 min	#16 of 31	Epoch AI Benchmarking Hub	2026-10-13
GPQA Diamond	67.6%	#148 of 275	Epoch AI Benchmarking Hub	2026-10-13
DTBench	64.8%	#84 of 146	Epoch AI Benchmarking Hub	2026-10-13
OTIS Mock AIME	37.8%	#115 of 189	Epoch AI Benchmarking Hub	2026-10-13
WeirdML	36.1%	#77 of 109	Epoch AI Benchmarking Hub	2026-10-13
Fiction.liveBench	50.0%	#34 of 47	Epoch AI Benchmarking Hub	2026-10-13
SciCode	35.8%	#78 of 106	Epoch AI Benchmarking Hub	2026-10-13
SimpleBench	27.2%	#46 of 61	Epoch AI Benchmarking Hub	2026-10-13
CritPt	0.0%	#88 of 115	Epoch AI Benchmarking Hub	2026-10-13
LMCA	15.5%	#86 of 108	Epoch AI Benchmarking Hub	2026-10-13

About DeepSeek-V3 (Mar 2025)

DeepSeek-V3 (Mar 2025) is an open-weight model that can run on your own hardware from DeepSeek. It was released on 2025-03-24, according to Epoch AI. The cheapest API listing we track charges $0.200 per million input tokens and $0.200 per million output tokens, through hyperbolic. We hold 11 benchmark results for it across 4 areas, and none is in the top tenth of its board.

Where DeepSeek-V3 (Mar 2025) stands

On our aggregate score, DeepSeek-V3 (Mar 2025) is #647 of 743 ranked models, in the lower half. Its score is 44.1; the scale is centred on 50, so it sits 5.9 points below the typical model. That beats 13% of the ranked field. Among the 33 ranked DeepSeek models it is #31. The next one up is DeepSeek-V3.2 (Thinking; Novita) at 44.5. The next one down is deepseek-r1-0528-qwen3-8b at 43.6.

DeepSeek-V3 (Mar 2025), benchmark by benchmark

Epoch ECI puts DeepSeek-V3 (Mar 2025) at 136, #68 of 204. That beats 67% of the models tested, but 20 points short of Qwen 3.8 Max (156). The results either side are qwen3-8b (136) and DeepSeek-R1-Distill-Qwen-14B (135). It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so DeepSeek-V3 (Mar 2025) is 9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on METR time horizon: 23 min, #16 of 31, with 48% of the field scoring higher. Claude Mythos Preview (Early) leads at 1,045 min; the closest result above is DeepSeek-R1 at 27 min. It covers length of task (minutes) done at 50% success. The median score on this board is also 23 min. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on GPQA Diamond: 67.6%, #148 of 275, with 53% of the field scoring higher. GPT-6 Astra (max) leads at 95.8%; the closest result above is o3-mini-2025-01-31 (low) at 68.2%. It covers expert-level science questions. The board median is 71.7%, so DeepSeek-V3 (Mar 2025) is 4.1 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on DTBench: 64.8%, #84 of 146, with 57% of the field scoring higher. Claude Opus 5.5 (max) leads at 98.9%; the closest result above is Mistral Large 3 at 65.1%. It covers dynamic tool-use benchmark. The board median is 68.8%, so DeepSeek-V3 (Mar 2025) is 4.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on OTIS Mock AIME: 37.8%, #115 of 189, with 60% of the field scoring higher. Claude Fable 5.1 (max) leads at 100.0%; the closest result above is GPT-4.1 at 38.3%. It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so DeepSeek-V3 (Mar 2025) is 20.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on WeirdML: 36.1%, #77 of 109, with 70% of the field scoring higher. GPT-6 Astra (max) leads at 93.3%; the closest result above is DeepSeek-R1 at 36.5%. It covers unusual machine-learning tasks. The board median is 41.2%, so DeepSeek-V3 (Mar 2025) is 5.1 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on SciCode: 35.8%, #78 of 106, with 73% of the field scoring higher. Claude Opus 5.5 (max) leads at 66.9%; the closest result above is Nemotron 3 Super at 36.0%. It covers scientific research code. The board median is 44.3%, so DeepSeek-V3 (Mar 2025) is 8.6 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on SimpleBench: 27.2%, #46 of 61, with 74% of the field scoring higher. Claude Fable 5 (max) leads at 81.9%; the closest result above is Claude 3.5 Sonnet (Jun 2024) at 27.5%. It covers trick questions that test common sense. The board median is 41.4%, so DeepSeek-V3 (Mar 2025) is 14.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on CritPt: 0.0%, #88 of 115, with 76% of the field scoring higher. GPT-5.6 Sol (max) leads at 32.3%; the closest result above is DeepSeek-V3 at 0.0%. It covers unpublished research-level physics problems. The board median is 2.9%, so DeepSeek-V3 (Mar 2025) is 2.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

DeepSeek-V3 (Mar 2025) lands below the middle on LMCA: 15.5%, #86 of 108, with 79% of the field scoring higher. Claude Opus 5.5 (max) leads at 68.2%; the closest result above is Qwen3-30B-A3B at 15.8%. It covers language-model capability assessment. The board median is 29.2%, so DeepSeek-V3 (Mar 2025) is 13.8 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

The full table above also lists the other 1 result, including boards with fewer than ten models, where a rank says little.

Strengths and weaknesses

Reasoning, maths and science: uneven, ahead of 33% of the field on average across 4 benchmarks. Its best board here is GPQA Diamond (#148 of 275) and its weakest is CritPt (#88 of 115). General knowledge and chat: uneven, ahead of 38% of the field on average across 5 benchmarks. Its best board here is Epoch ECI (#68 of 204) and its weakest is LMCA (#86 of 108).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so DeepSeek-V3 (Mar 2025) 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 MiniMax-M3 ($0.230) and MiniMax-M2.7 ($0.250).

Release history

The DeepSeek model released before it was DeepSeek V3 0324 on 2025-03-24, which scores 50.8. The next one was deepseek-r1-0528-qwen3-8b on 2025-05-28, at 43.6.

Reading these numbers

The 11 scores on this page come from 1 source: Epoch AI Benchmarking Hub (11). No official leaderboard supplied any of them, so treat them as third-party measurements.

Models ranked nearby

#644 gpt-oss-20b (high) (44.3)
#645 Yi-6B-Chat (44.3)
#646 Claude 3.5 Sonnet (Jun 2024) (44.2)
#648 Yi-34B-Chat (44.0)
#649 o3-mini (high) (44.0)
#650 gemini-1.5-pro-001 (44.0)