Chinese AI lab DeepSeek’s last model release was V3.2 (and V3.2 Speciale) last December (https://simonwillison.net/2025/Dec/1/deepseek-v32/). They just dropped the first of their hotly anticipated V4 series in the shape of two preview models, DeepSeek-V4-Pro (https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro) and DeepSeek-V4-Flash (https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash).

Both models are 1 million token context Mixture of Experts. Pro is 1.6T total parameters, 49B active. Flash is 284B total, 13B active. They’re using the standard MIT license.

I think this makes DeepSeek-V4-Pro the new largest open weights model. It’s larger than Kimi K2.6 (1.1T) and GLM-5.1 (754B) and more than twice the size of DeepSeek V3.2 (685B).

Pro is 865GB on Hugging Face, Flash is 160GB. I’m hoping that a lightly quantized Flash will run on my 128GB M5 MacBook Pro. It’s possible the Pro model may run on it if I can stream just the necessary active experts from disk.

For the moment I tried the models out via OpenRouter (https://openrouter.ai/), using llm-openrouter (https://github.com/simonw/llm-openrouter):

llm install llm-openrouter llm openrouter refresh llm -m openrouter/deepseek/deepseek-v4-pro ‘Generate an SVG of a pelican riding a bicycle’

Here’s the pelican for DeepSeek-V4-Flash (https://gist.github.com/simonw/4a7a9e75b666a58a0cf81495acddf529):

And for DeepSeek-V4-Pro (https://gist.github.com/simonw/9e8dfed68933ab752c9cf27a03250a7c):

For comparison, take a look at the pelicans I got from DeepSeek V3.2 in December (https://simonwillison.net/2025/Dec/1/deepseek-v32/), V3.1 in August (https://simonwillison.net/2025/Aug/22/deepseek-31/), and V3-0324 in March 2025 (https://simonwillison.net/2025/Mar/24/deepseek/).

So the pelicans are pretty good, but what’s really notable here is the cost. DeepSeek V4 is a very, very inexpensive model.

Here’s DeepSeek’s pricing page (https://api-docs.deepseek.com/quick_start/pricing). They’re charging $0.14/million tokens input and $0.28/million tokens output for Flash, and $1.74/million input and \(3.48/million output for Pro. Here's a comparison table with the frontier models from Gemini, OpenAI and Anthropic: Model Input (\)/M) Output ($/M)

DeepSeek V4 Flash $0.14 $0.28

GPT-5.4 Nano $0.20 $1.25

Gemini 3.1 Flash-Lite $0.25 $1.50

Gemini 3 Flash Preview $0.50 $3

GPT-5.4 Mini $0.75 $4.50

Claude Haiku 4.5 $1 $5

DeepSeek V4 Pro $1.74 $3.48

Gemini 3.1 Pro $2 $12

GPT-5.4 $2.50 $15

Claude Sonnet 4.6 $3 $15

Claude Opus 4.7 $5 $25

GPT-5.5 $5 $30

DeepSeek-V4-Flash is the cheapest of the small models, beating even OpenAI’s GPT-5.4 Nano. DeepSeek-V4-Pro is the cheapest of the larger frontier models.

This note from the DeepSeek paper (https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash/blob/main/DeepSeek_V4.pdf) helps explain why they can price these models so low - they’ve focused a great deal on efficiency with this release, especially for longer context prompts:

In the scenario of 1M-token context, even DeepSeek-V4-Pro, which has a larger number of activated parameters, attains only 27% of the single-token FLOPs (measured in equivalent FP8 FLOPs) and 10% of the KV cache size relative to DeepSeek-V3.2. Furthermore, DeepSeek-V4-Flash, with its smaller number of activated parameters, pushes efficiency even further: in the 1M-token context setting, it achieves only 10% of the single-token FLOPs and 7% of the KV cache size compared with DeepSeek-V3.2.

DeepSeek’s self-reported benchmarks in their paper (https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash/blob/main/DeepSeek_V4.pdf) show their Pro model competitive with those other frontier models, albeit with this note:

Through the expansion of reasoning tokens, DeepSeek-V4-Pro-Max demonstrates superior performance relative to GPT-5.2 and Gemini-3.0-Pro on standard reasoning benchmarks. Nevertheless, its performance falls marginally short of GPT-5.4 and Gemini-3.1-Pro, suggesting a developmental trajectory that trails state-of-the-art frontier models by approximately 3 to 6 months.

I’m keeping an eye on huggingface.co/unsloth/models (https://huggingface.co/unsloth/models) as I expect the Unsloth team will have a set of quantized versions out pretty soon. It’s going to be very interesting to see how well that Flash model runs on my own machine.

    Tags: ai (https://simonwillison.net/tags/ai), generative-ai (https://simonwillison.net/tags/generative-ai), llms (https://simonwillison.net/tags/llms), llm (https://simonwillison.net/tags/llm), llm-pricing (https://simonwillison.net/tags/llm-pricing), pelican-riding-a-bicycle (https://simonwillison.net/tags/pelican-riding-a-bicycle), deepseek (https://simonwillison.net/tags/deepseek), llm-release (https://simonwillison.net/tags/llm-release), openrouter (https://simonwillison.net/tags/openrouter), ai-in-china (https://simonwillison.net/tags/ai-in-china)