DeepSeek-V4-Pro-Max: Benchmarks, Pricing & Context Window
DeepSeek-V4-Pro-Max is a language model from DeepSeek, released in April 2026, with a 1.0M-token context window, and pricing from $1.30/M input, $0.100/M cached input, $2.60/M output.
DeepSeek-V4-Pro-Max is the maximum reasoning effort mode of DeepSeek-V4-Pro, a 1.6T-parameter MoE model with 49B activated parameters and a 1M-token context window. It introduces a hybrid attention architecture combining Compressed Sparse
DeepSeek-V4-Pro-Max benchmarks
Capability tiers
Standing within each category, adjusted for leaderboard depth.
Real tasks performance
High-confidence performance for DeepSeek-V4-Pro-Max across real-world prompt categories. Only 95% intervals at most 4 points wide are shown.
Performance by conversation depth
How DeepSeek-V4-Pro-Max holds up as conversations get longer.
Quality Tracker
DeepSeek-V4-Pro-Max Performance Across Datasets
Scores sourced from the model's scorecard, paper, or official blog posts
llm-stats.com - Wed Sep 09 2026DeepSeek-V4-Pro-Max pricing
Providers
DeepSeek-V4-Pro-Max starts at $1.30 per million input tokens and $2.60 per million output tokens via DeepInfra. Reused prompt prefixes cost $0.100 per million cached input tokens. See all 5 providers below with their per-token pricing, latency, throughput, and modality support.
| Provider | Input $/M | Cached input $/M | Output $/M | Context in / out | TTFT p95 s | Output p5 c/s | Modalities in / out |
|---|---|---|---|---|---|---|---|
 DeepInfra | $1.30 | $0.100 | $2.60 | 1.0M/1.0M | — | — | / |
 Novita | $1.60 | $0.135 | $3.20 | 1.0M/131.1K | — | — | / |
 DeepSeek | $1.74 | — | $3.48 | 1.0M/393.2K | — | — | / |
 Fireworks | $1.74 | — | $3.48 | 1.0M/65.5K | — | — | / |
 Together | $1.74 | — | $3.48 | 524.3K/65.5K | — | — | / |
Cached input is the discounted price for prompt tokens served from a provider cache. TTFT is time to first token. Output is characters per second; p5 is the sustained floor exceeded by 95% of observed requests.
DeepSeek-V4-Pro-Max model size
DeepSeek-V4-Pro-Max has 1.6 trillion parameters and was trained on 32 trillion tokens. See how it compares to other models in the same parameter range.
DeepSeek-V4-Pro-Max context window
Input and output token limits for DeepSeek-V4-Pro-Max, plus how it ranks on long-context understanding.
Try now
huggleMake it with
DeepSeek-V4-Pro-Max.
DeepSeek-V4-Pro-Max latency
DeepSeek-V4-Pro-Max time to first token, sustained output throughput, and failed-request rate from live model usage over the trailing 7 days.
DeepSeek-V4-Pro-Max examples
Recent arena outputs from DeepSeek-V4-Pro-Max, picked from the highest-ranked matchups.
DeepSeek-V4-Pro-Max license
DeepSeek-V4-Pro-Max is released under the MIT license, which permits commercial use, has 1.6T parameters.
- License
- MIT
- Commercial use allowed
- Parameters
- 1.6T
MIT License - allows commercial use
DeepSeek-V4-Pro-Max resources
Official sources for DeepSeek-V4-Pro-Max: provider documentation, official playground, source repository, model weights.
DeepSeek-V4-Pro-Max vs other models
The most-compared alternatives to DeepSeek-V4-Pro-Max are Claude Opus 4.6, Claude Opus 4.5, ERNIE 5.0. Open any pair side-by-side for benchmarks, pricing, context, and latency.
Models like DeepSeek-V4-Pro-Max
Models ranked just above and below DeepSeek-V4-Pro-Max by LLM Stats score.
FAQ
Common questions about DeepSeek-V4-Pro-Max.