DeepSeek V3 costs $0.32 per 1M input tokens and $0.89 per 1M output tokens on the standard tier. It holds 163,840 tokens of context.On the atlas: grid D3. 10 other models share the square; the cheapest is Codestral 2508 at $0.30 per 1M input.
Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
Cite
Cite this page
ModelsAtlas. “DeepSeek V3”. Retrieved 10 Oct 2026. https://modelsatlas.com/models/deepseek/v3
- Input / 1M
$0.32
observed We measured- Source
- Our catalogue
- Retrieved
- 10 Oct 2026
- Output / 1M
$0.89
observed We measured- Source
- Our catalogue
- Retrieved
- 10 Oct 2026
- Context
- tokens
164K
observed We measured- Source
- Our catalogue
- Retrieved
- 1 Sep 2026
- Offers
- 1 offer
Standard tier
Standard tier
≈ 123,000 words
1 host · standard
Hosts and pricing
1 offer · tiers are rows, not pagesSuggest a correction
| Host | Tier | Input / 1M | Output / 1M |
|---|---|---|---|
| Our cataloguechecked 10 Oct 2026 | standard | $0.32observed We measured
| $0.89observed We measured
|
Paying too much? Models near DeepSeek V3 on the atlas.Cheaper alternatives to DeepSeek V3 →
What it costs for your workload
Specifications
Tap src for a sourceSuggest a correction
| Catalogue id | deepseek-v3 | |
|---|---|---|
| Creator | DeepSeek | |
| Context window | 163,840 tokens | Observed |
| Modalities | Not disclosed | |
| Licence | Not disclosed | |
| Weights | Not disclosed | |
| Variants | One (base) | |
| Record checked | 2026-09-01 |
Capabilities
“Not disclosed” means our records do not say, not that the model lacks it.
- Vision inputNot disclosed
- AudioNot disclosed
- Open weightsNot disclosed
- Free tierNone listed
- Batch tierNone listed
We hold no data yet on tool calling, structured output or benchmark scores, so none are shown.
Cheaper alternatives to DeepSeek V3
All 15 alternatives →| Model | Grid | Input / 1M | Output / 1M | Context | |
|---|---|---|---|---|---|
| Codestral 2508Mistral AI | GRIDD3 | $0.30 | $0.90 | 256K | Compare |
| Command A+Cohere | GRIDD3 | $0.30 | $1.50 | 192K | Compare |
| MiniMax M2MiniMax | GRIDD3 | $0.30 | $1.20 | 205K | Compare |
| MiniMax M2.1MiniMax | GRIDD3 | $0.30 | $1.20 | 205K | Compare |
| Muse Glimmer 30BMeta | GRIDD3 | $0.30 | $1.20 | 131K | Compare |
Ranked by distance from grid D3, then price.See all 15 alternatives →
Head-to-head comparisons
Works with
Where DeepSeek V3 connects across the atlas
◆Hosts serving it
- Our catalogue1 offer
◉Tools built on it
- None linked yet. Tell us about one
Questions about DeepSeek V3
How much does DeepSeek V3 cost?
On the cheapest standard-tier offer in our catalogue, input costs $0.32 per 1M tokens and output costs $0.89 per 1M tokens.
What is the DeepSeek V3 context window?
163,840 tokens, roughly 123,000 English words in one request.
Is there a free tier for DeepSeek V3?
No free-tier offer is listed for DeepSeek V3 in our catalogue. Its cheapest standard tier is $0.32 per 1M input tokens.
What is the cheapest alternative to DeepSeek V3?
Among the models nearest it on the atlas, Ministral 3 3B 2512 is the cheapest: $0.10 per 1M input tokens, 69% less, with 131K of context.
What does grid D3 mean for DeepSeek V3?
Grid D3 is input $0.30–1 per 1M tokens with context 128–256K. 10 other models share that square: Codestral 2508, Command A+, MiniMax M2 and 7 more. It describes price and context, not quality.
Sources
- Our catalogue — context window, modalities, licenceObserved · 1 Sep 2026
- Our catalogue — prices and tiersObserved · 10 Oct 2026
- Our calculation — grid reference, monthly cost and percentage comparisonsDerived · 10 Oct 2026
Spotted something wrong? Report a correction · Methodology
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- How much does DeepSeek V3 cost?
- What is the DeepSeek V3 context window?
- Is there a free tier for DeepSeek V3?