FRENES

DeepSeek · language model

DeepSeek V3: pricing, context and benchmarks

DeepSeek V3 is DeepSeek's model released on December 26, 2024. It costs $0.26 per million input tokens and $1.03 per million output tokens, with a 163.8K-token context window. Epoch AI ranks it #113 of 125 rated models.

Mid-range costOpen weightsTool calling

What you need to know about DeepSeek V3

DeepSeek, based in China, put DeepSeek V3 online on December 26, 2024. It accepts text as input and produces text. Its context window reaches 163,840 tokens, larger than 32 % of the models in the registry, and it can write up to 16,000 tokens in one go.

Released more than 1 year ago, it belongs to a generation most vendors have since superseded.

DeepSeek V3 pricing

On OpenRouter, DeepSeek V3 costs $0.26 per million input tokens and $1.03 per million output tokens. Counting three tokens read for every token written, its average price is $0.45 per million, more expensive than 39 % of paid models (1.7 times cheaper than the median of $0.76).

A concrete example: a task that sends 2,000 tokens (about 1,500 words) and receives a 500-token answer costs $0.001. At 1,000 tasks a day, budget $31 a month. The price calculator redoes this with your own volumes.

Performance and benchmarks

The Epoch Capabilities Index (ECI), which aggregates published results across many benchmarks, gives DeepSeek V3 a score of 132.3 (range 128.1 to 135.2). That places it #113 of 125 rated models.

Scores published by Epoch AI on the benchmarks the site tracks:

Source: Epoch AI, Benchmarking Hub (CC BY 4.0).
BenchmarkWhat it measuresScoreDate
GPQA DiamondPhD-level science questions71.2 %Dec 1, 2025
OTIS Mock AIMEmaths competition problems48.9 %Dec 1, 2025
Aider Polyglotcoding exercises in several languages70.2 %Dec 1, 2025
GPQA DiamondGPQA Diamond : 71.2 %71.2 %OTIS Mock AIMEOTIS Mock AIME : 48.9 %48.9 %Aider PolyglotAider Polyglot : 70.2 %70.2 %
DeepSeek V3 success rate by benchmark (Epoch AI).
116129142156169$0.1$1$10Average price ($ per million tokens, log scale)ECIClaude Opus 5.5 : Average price ($ per million tokens, log scale) $8, ECI 169GPT-6 Astra : Average price ($ per million tokens, log scale) $20, ECI 169GPT-6.1 Sol : Average price ($ per million tokens, log scale) $4, ECI 168Claude Fable 5.1 : Average price ($ per million tokens, log scale) $20, ECI 166Gemini 4 Argon : Average price ($ per million tokens, log scale) $4, ECI 166Claude Sonnet 5.5 : Average price ($ per million tokens, log scale) $4, ECI 166Claude Opus 5 : Average price ($ per million tokens, log scale) $10, ECI 164GPT-6 Sol : Average price ($ per million tokens, log scale) $4, ECI 164Claude Fable 5 : Average price ($ per million tokens, log scale) $20, ECI 163GPT-5.5 Pro : Average price ($ per million tokens, log scale) $68, ECI 163GPT-5.6 Sol : Average price ($ per million tokens, log scale) $8, ECI 163Muse Wonder : Average price ($ per million tokens, log scale) $8, ECI 162GPT-5.6 Terra : Average price ($ per million tokens, log scale) $5, ECI 160GPT-5.5 : Average price ($ per million tokens, log scale) $11, ECI 160GPT-5.4 Pro : Average price ($ per million tokens, log scale) $68, ECI 160Claude Haiku 5.5 : Average price ($ per million tokens, log scale) $0.2, ECI 159Claude Opus 4.8 : Average price ($ per million tokens, log scale) $10, ECI 159Gemini 3.7 Flash : Average price ($ per million tokens, log scale) $2, ECI 158Kimi K3 : Average price ($ per million tokens, log scale) $3, ECI 158Gemini 3.8 Flash : Average price ($ per million tokens, log scale) $2, ECI 157GPT-5.4 : Average price ($ per million tokens, log scale) $6, ECI 157Muse Spark 1.3 : Average price ($ per million tokens, log scale) $2, ECI 157GPT-5.3-Codex : Average price ($ per million tokens, log scale) $5, ECI 157Claude Sonnet 5 : Average price ($ per million tokens, log scale) $4, ECI 157Claude Opus 4.7 : Average price ($ per million tokens, log scale) $10, ECI 157GPT-6 Luna : Average price ($ per million tokens, log scale) $0.2, ECI 157Grok 4.6 : Average price ($ per million tokens, log scale) $3, ECI 157GPT-5.6 Luna : Average price ($ per million tokens, log scale) $0.5, ECI 157GPT-5.2 Pro : Average price ($ per million tokens, log scale) $58, ECI 156GLM 5.3 : Average price ($ per million tokens, log scale) $0.4, ECI 156DeepSeek V4.1 Flash 1017 : Average price ($ per million tokens, log scale) $0.3, ECI 156DeepSeek V4.1 Flash : Average price ($ per million tokens, log scale) $0.3, ECI 156Claude Opus 4.6 : Average price ($ per million tokens, log scale) $10, ECI 156DeepSeek V4 Pro 0813 : Average price ($ per million tokens, log scale) $1, ECI 156DeepSeek V4 Pro 0423 : Average price ($ per million tokens, log scale) $0.3, ECI 156Gemini 3.1 Pro Preview : Average price ($ per million tokens, log scale) $5, ECI 155Muse Spark Open : Average price ($ per million tokens, log scale) $2, ECI 155Qwen3.8 Max (0902) : Average price ($ per million tokens, log scale) $3, ECI 155Muse Spark 1.2 : Average price ($ per million tokens, log scale) $2, ECI 155Gemini 3.5 Flash : Average price ($ per million tokens, log scale) $3, ECI 155DeepSeek V4 Flash 0731 : Average price ($ per million tokens, log scale) $0.07, ECI 155DeepSeek V4 Flash 0423 : Average price ($ per million tokens, log scale) $0.1, ECI 155Gemini 3.6 Flash : Average price ($ per million tokens, log scale) $2, ECI 155Grok 4.7 : Average price ($ per million tokens, log scale) $3, ECI 154Grok 4.5 : Average price ($ per million tokens, log scale) $3, ECI 154Muse Spark 1.1 : Average price ($ per million tokens, log scale) $2, ECI 154Qwen3.7 Max : Average price ($ per million tokens, log scale) $2, ECI 154GPT-5.2 : Average price ($ per million tokens, log scale) $5, ECI 154GLM 5.3 Flash : Average price ($ per million tokens, log scale) $0.2, ECI 153Claude Sonnet 4.6 : Average price ($ per million tokens, log scale) $6, ECI 153Grok 4.20 : Average price ($ per million tokens, log scale) $2, ECI 152GLM 5.2 : Average price ($ per million tokens, log scale) $2, ECI 152Gemini 3 Flash Preview : Average price ($ per million tokens, log scale) $1, ECI 152Qwen3.8 27B : Average price ($ per million tokens, log scale) $0.8, ECI 152Kimi K2.6 : Average price ($ per million tokens, log scale) $1, ECI 151GPT-5 Pro : Average price ($ per million tokens, log scale) $41, ECI 151Inkling Small : Average price ($ per million tokens, log scale) $0.6, ECI 150Kimi K2.7 Code : Average price ($ per million tokens, log scale) $1, ECI 150Claude Opus 4.5 : Average price ($ per million tokens, log scale) $10, ECI 150Mistral Large 4 : Average price ($ per million tokens, log scale) $2, ECI 150GPT-5 : Average price ($ per million tokens, log scale) $3, ECI 150GLM 5.1 : Average price ($ per million tokens, log scale) $2, ECI 150GPT-5.1 : Average price ($ per million tokens, log scale) $3, ECI 150GPT-5.4 Mini : Average price ($ per million tokens, log scale) $2, ECI 149Inkling : Average price ($ per million tokens, log scale) $2, ECI 148Kimi K2.5 : Average price ($ per million tokens, log scale) $1, ECI 148o3 Pro : Average price ($ per million tokens, log scale) $35, ECI 147Qwen3.6 Plus : Average price ($ per million tokens, log scale) $0.7, ECI 147Qwen3.7 Plus : Average price ($ per million tokens, log scale) $0.6, ECI 147MiniMax M3 : Average price ($ per million tokens, log scale) $0.5, ECI 147Qwen3.5 397B A17B : Average price ($ per million tokens, log scale) $1, ECI 147Claude Sonnet 4.5 : Average price ($ per million tokens, log scale) $6, ECI 147o3 : Average price ($ per million tokens, log scale) $4, ECI 147MiniMax M2.5 : Average price ($ per million tokens, log scale) $0.5, ECI 147Qwen3.6 27B : Average price ($ per million tokens, log scale) $0.7, ECI 146DeepSeek V3.2 : Average price ($ per million tokens, log scale) $0.3, ECI 146GPT-5.4 Nano : Average price ($ per million tokens, log scale) $0.5, ECI 146Nemotron 3 Ultra : Average price ($ per million tokens, log scale) $1, ECI 146Kimi K2 Thinking : Average price ($ per million tokens, log scale) $1, ECI 146GLM 5 : Average price ($ per million tokens, log scale) $2, ECI 146MiniMax M2.7 : Average price ($ per million tokens, log scale) $0.5, ECI 146o4 Mini : Average price ($ per million tokens, log scale) $2, ECI 145GPT-5 Mini : Average price ($ per million tokens, log scale) $0.7, ECI 145Gemini 3.5 Flash Lite : Average price ($ per million tokens, log scale) $0.8, ECI 145DeepSeek V3.2 Exp : Average price ($ per million tokens, log scale) $0.3, ECI 145Gemini 3.1 Flash Lite : Average price ($ per million tokens, log scale) $0.6, ECI 144Gemini 3.1 Flash Lite Preview : Average price ($ per million tokens, log scale) $0.6, ECI 144Claude Opus 4.1 : Average price ($ per million tokens, log scale) $30, ECI 144Qwen3.7 Flash : Average price ($ per million tokens, log scale) $0.06, ECI 144Qwen3.6 35B A3B : Average price ($ per million tokens, log scale) $0.3, ECI 144GLM 4.7 : Average price ($ per million tokens, log scale) $0.9, ECI 144Gemma 4 31B : Average price ($ per million tokens, log scale) $0.3, ECI 143Qwen3.6 Flash : Average price ($ per million tokens, log scale) $0.4, ECI 143Qwen3.5-35B-A3B : Average price ($ per million tokens, log scale) $0.4, ECI 142Claude Haiku 4.5 : Average price ($ per million tokens, log scale) $2, ECI 142Gemma 4 26B A4B : Average price ($ per million tokens, log scale) $0.1, ECI 142o1 : Average price ($ per million tokens, log scale) $26, ECI 142Mistral Medium 3.5 : Average price ($ per million tokens, log scale) $3, ECI 141o3 Mini : Average price ($ per million tokens, log scale) $2, ECI 140DeepSeek V3.1 : Average price ($ per million tokens, log scale) $0.4, ECI 140gpt-oss-120b : Average price ($ per million tokens, log scale) $0.3, ECI 140Qwen3.5-9B : Average price ($ per million tokens, log scale) $0.2, ECI 139GPT-5 Nano : Average price ($ per million tokens, log scale) $0.1, ECI 139R1 0528 : Average price ($ per million tokens, log scale) $0.9, ECI 139Qwen3 32B : Average price ($ per million tokens, log scale) $0.1, ECI 138Qwen3 14B : Average price ($ per million tokens, log scale) $0.1, ECI 138gpt-oss-20b : Average price ($ per million tokens, log scale) $0.04, ECI 138GPT-4.1 : Average price ($ per million tokens, log scale) $4, ECI 137Qwen3 30B A3B : Average price ($ per million tokens, log scale) $0.2, ECI 136GPT-4.1 Mini : Average price ($ per million tokens, log scale) $0.7, ECI 135Mistral Medium 3 : Average price ($ per million tokens, log scale) $0.8, ECI 134DeepSeek V3 0324 : Average price ($ per million tokens, log scale) $0.5, ECI 132Llama 4 Maverick : Average price ($ per million tokens, log scale) $0.3, ECI 132Phi 4 : Average price ($ per million tokens, log scale) $0.09, ECI 130Gemma 3 27B : Average price ($ per million tokens, log scale) $0.1, ECI 130Llama 4 Scout : Average price ($ per million tokens, log scale) $0.1, ECI 130GPT-4.1 Nano : Average price ($ per million tokens, log scale) $0.2, ECI 130GPT-4o-mini : Average price ($ per million tokens, log scale) $0.3, ECI 126GPT-4o-mini (2024-07-18) : Average price ($ per million tokens, log scale) $0.3, ECI 126Gemma 3 12B : Average price ($ per million tokens, log scale) $0.07, ECI 124Gemma 2 27B : Average price ($ per million tokens, log scale) $0.7, ECI 122Mistral Large : Average price ($ per million tokens, log scale) $3, ECI 122Mistral Nemo : Average price ($ per million tokens, log scale) $0.03, ECI 119Gemma 3 4B : Average price ($ per million tokens, log scale) $0.04, ECI 116DeepSeek V3 : Average price ($ per million tokens, log scale) $0.5, ECI 132DeepSeek V3Claude Opus 5.5
Each dot is a model rated by Epoch AI; DeepSeek V3 is in orange. Higher = more capable, further left = cheaper. Open the LLM benchmark to filter.

For at least equal performance, Claude Haiku 5.5, GPT-6 Luna, and DeepSeek V4.1 Flash 1017 cost at least 30% less.

Features, privacy and local use

  • It supports tool calling, so it can drive an AI agent, call APIs or MCP servers.
  • It can answer in an enforced JSON format (structured outputs).
  • Its knowledge stops at July 2024.

Its weights are open and downloadable from Hugging Face: you can run it on your own machines without sending your data to a third party. With 684,489,845,504 billion parameters, it needs about 382,488,507,128 GB of memory at 4-bit quantisation. Check with the VRAM calculator. It was downloaded 1,420,996 times on Hugging Face last month.

Other DeepSeek models

Average price per million tokens (3 input for 1 output).
ModelReleasedAvg. priceECI
DeepSeek V4.1 Flash 1017Oct 17, 2026$0.26155.9
DeepSeek V4.1 FlashSep 10, 2026$0.27155.9
DeepSeek V4 Flash Vision ExpAug 21, 2026$0.32—
DeepSeek V4 Pro 0813Aug 12, 2026$0.99155.6
DeepSeek V4 Flash 0731Jul 31, 2026$0.07154.8
DeepSeek V4 Pro 0423Apr 24, 2026$0.26155.6
DeepSeek V4 Flash 0423Apr 24, 2026$0.11154.8
DeepSeek V3.2Dec 1, 2025$0.32146.1

Verdict

Computed score: 39/100, from public data (performance, price, context, features).

Strengths

  • Open weights: self-hostable, your data stays with you.
  • Agent-ready: tool calling and structured JSON output.

Limitations

  • Modest performance: behind 89 % of rated models.
  • Older generation, often replaced by a newer model from the same vendor.
  • Text only: reads neither images nor files.

Avoid if…

  • you want to host it on a single consumer computer (≈ 382,488,507,128 GB needed at 4-bit).

Alternatives

Comparisons and tools

Learn more

Frequently asked questions

How much does DeepSeek V3 cost?

$0.26 per million input tokens and $1.03 per million output tokens on OpenRouter (checked October 25, 2026). A task with 2,000 tokens read and 500 written costs $0.001.

How big is DeepSeek V3's context window?

163,840 tokens, roughly 123,000 words, with a maximum of 16,000 output tokens.

Can DeepSeek V3 run locally?

Yes, its weights are published. Expect about 382,488,507,128 GB of memory at 4-bit.

Is DeepSeek V3 suitable for an AI agent?

Technically yes: it supports tool calling and structured outputs. Then check its cost for your volume.

How good is DeepSeek V3?

Epoch AI ranks it #113 of 125 rated models (ECI 132.3). On GPQA Diamond, it scores 71.2%.

What are the alternatives to DeepSeek V3?

The closest models on price and performance are Mistral Medium 3, GPT-4o-mini, and GPT-4o-mini (2024-07-18).

Related articles

Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score