DeepSeek-V3.2-Speciale: benchmark scores, price and ranking

DeepSeek-V3.2-Speciale from DeepSeek ranks #195 of 743 LLMs on the LLMs Tiger aggregate benchmark score (50.4), based on 2 benchmark results captured up to 2026-10-13.

Organisation: DeepSeek
Type: Open-weight / local model
Open weights: Yes
Released: 2025-12-01 (source: epoch)
Cheapest API price: $0.580 in / $1.68 out per million tokens (via azure_ai, observed 2026-10-13T20:47Z)
Aggregate rank: #195 of 743 (score 50.4)

Across the 2 benchmarks where at least ten models are scored, DeepSeek-V3.2-Speciale is in the top 10% on 0 and in the top half on 2.

Its cheapest listed input price, $0.580 per million tokens, is lower than that of 46% of the 427 models with a listed price.

DeepSeek-V3.2-Speciale benchmark results

DeepSeek-V3.2-Speciale benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
SimpleBench	52.6%	#19 of 61	Epoch AI Benchmarking Hub	2026-10-13
WeirdML	46.7%	#34 of 109	Epoch AI Benchmarking Hub	2026-10-13

About DeepSeek-V3.2-Speciale

DeepSeek-V3.2-Speciale is an open-weight model that can run on your own hardware from DeepSeek. It was released on 2025-12-01, according to Epoch AI. The cheapest API listing we track charges $0.580 per million input tokens and $1.68 per million output tokens, through azure_ai. We hold 2 benchmark results for it across 2 areas, and none is in the top tenth of its board.

Where DeepSeek-V3.2-Speciale stands

On our aggregate score, DeepSeek-V3.2-Speciale is #195 of 743 ranked models, around the middle of the table. Its score is 50.4; the scale is centred on 50, so it sits 0.4 points above the typical model. That beats 74% of the ranked field. Among the 33 ranked DeepSeek models it is #10. The next one up is DeepSeek Chat V3 (prev) at 50.5. The next one down is DeepSeek v4 (unknown) at 50.0.

DeepSeek-V3.2-Speciale, benchmark by benchmark

SimpleBench puts DeepSeek-V3.2-Speciale at 52.6%, #19 of 61. That beats 69% of the models tested, but 29.3 points short of Claude Fable 5 (max) (81.9%). The results either side are o3 (high) (53.1%) and Gemini 2.5 Pro Exp (Mar 2025) (51.6%). It covers trick questions that test common sense. The board median is 41.4%, so DeepSeek-V3.2-Speciale is 11.2 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On WeirdML, DeepSeek-V3.2-Speciale reaches 46.7%. It ranks #34 of 109, ahead of 69% of the field, and 46.6 points behind GPT-6 Astra (max) at 93.3%. The results either side are o1-preview (47.6%) and claude-sonnet-4-20250514_16K (46.1%). It covers unusual machine-learning tasks. The board median is 41.2%, so DeepSeek-V3.2-Speciale is 5.6 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Strengths and weaknesses

DeepSeek-V3.2-Speciale has too few results on boards with ten or more models to say where it is strong or weak. Treat any single score as a data point, not a verdict.

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so DeepSeek-V3.2-Speciale costs more 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 Flash 0731 (max) ($0.016, 53.1), GLM-5.3-Flash (max) ($0.026, 49.6). Similar scores cost more on GLM-5 ($0.600) and kimi-k2p5 ($0.600).

Release history

The DeepSeek model released before it was DeepSeek-V3.2 on 2025-12-01, which scores 47.7. The next one was DeepSeek v4 Flash (max) on 2026-04-24, at 49.6.

Reading these numbers

The 2 scores on this page come from 1 source: Epoch AI Benchmarking Hub (2). No official leaderboard supplied any of them, so treat them as third-party measurements. With 2 results the aggregate rank can move a lot when one new score arrives.

Models ranked nearby

#192 ollama/Qwen2.5.1-Coder-7B-Instruct-GGUF:Q8_0-32k (50.4)
#193 Claude 3.7 Sonnet (64k thinking) (50.4)
#194 Nemotron-Cascade-2-30B-A3B (50.4)
#196 gemini-2.5-flash-preview-04-17 (50.4)
#197 Claude 4.5 Haiku (50.3)
#198 claude-opus-4-5-20251101-thinking-64k-high-effort (50.3)