The arena · AI model review
DeepSeek-V4
by DeepSeek
Open-weights 1.6T MoE with 1M-token context, now served as V4-Pro-0813 at $3.96 per 1M output tokens at peak
$3.96/1M out at peak, $1.98 off-peak
Official DeepSeek API price for deepseek-v4-pro (DeepSeek-V4-Pro-0813): peak $1.32/1M input (cache hit $0.044), $3.96/1M output; off-peak half price ($0.66/$1.98, cache hit $0.022). Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday. V4-Flash is retired: deepseek-v4-flash now routes to V4.1-Flash at Flash prices. Verified against api-docs.deepseek.com 2026-11-10.
DeepSeek
1M tokens (384K max output)
$1.32/1M in at peak, $0.66 off-peak (cache hit from $0.022)
$3.96/1M out at peak, $1.98 off-peak
text only
Yes
What is DeepSeek-V4?
Flagship open-weights MoE family from DeepSeek (V4-Pro 1.6T params/49B active, V4-Flash 284B/13B), released as a preview on April 24, 2026. Brings a 1M-token context window via sparse attention, thinking and non-thinking modes, and MIT-licensed weights on Hugging Face. Positioned as open-source SOTA on agentic coding benchmarks at a fraction of frontier API prices.
DeepSeek-V4 pros & cons
Pros
- 1M-token context window (8x the 128K of V3.2) with up to 384K output tokens, standard on the official API
- Still far below closed frontier prices: V4-Pro-0813 costs $1.32/$3.96 per 1M tokens at peak and $0.66/$1.98 off-peak, with cache hits from $0.022
- MIT-licensed open weights for both V4-Pro and V4-Flash on Hugging Face: commercial use, fine-tuning and redistribution allowed
- Open-source SOTA on agentic coding: 80.6 on SWE-bench Verified (Think Max config), tied with Gemini 3.1 Pro, plus Codeforces rating 3206 (~rank 23 vs humans)
- Ranks #3 of 93 on the Artificial Analysis Intelligence Index (score 44), well above the 25 average
- Sparse-attention stack cuts 1M-context inference to 27% of V3.2's FLOPs and 10% of its KV cache
Cons
- Intermittent malformed tool calls: function calls sometimes emitted as plain text in content instead of the tool_calls field (GitHub issue deepseek-ai #1244)
- Thinking mode breaks long multi-turn tool-call chains with 400 errors in agent frameworks (OpenClaw issue #72044, fix still incomplete)
- Developers report it fabricating nonexistent APIs in custom codebases and acting on hallucinated user input in agent loops
- Very verbose (180M eval output tokens vs 95M median) and mid-pack speed at 54.6 tok/s (#39/93), which erodes the low per-token price in practice
- Text only on the API (no vision, unlike V4.1-Flash), and peak-hour pricing doubles rates on weekdays at 01:00-04:00 and 06:00-10:00 UTC; V4-Flash is retired and its API names now route to V4.1-Flash
The arena’s verdict
Choose DeepSeek-V4 (now V4-Pro-0813 on the API) if you want near-frontier reasoning with MIT-licensed weights for self-hosting and fine-tuning, at prices still well below Claude Opus or GPT-6. But check DeepSeek-V4.1-Flash first: DeepSeek says it beats V4-Pro on agentic coding, it reads images, and it costs about a third of the price. Avoid V4 for production agents that depend on rock-solid multi-turn tool calling, where users have reported malformed tool calls and fabricated APIs, and for any vision or audio work. Budget for the peak-hour rate, which doubles the bill during Beijing business hours.
Thumbs up or thumbs down
Cast your verdict
Would you recommend DeepSeek-V4, or warn the crowd away?
Top DeepSeek-V4 alternatives
All alternativesGoogle workhorse September 2026: 1M context, $0.75/$3.75 promo pricing, 305 tokens/s output speed
Google's Nov 2025 frontier model: 1M context, first past 1500 Elo on LMArena, replaced by Gemini 3.1 Pro.
Compare DeepSeek-V4 head-to-head
What the crowd says
“Tool calling is flaky. Function calls sometimes land as plain text instead of the tool_calls field, and thinking mode 400s on long multi-turn chains. Not agent-ready yet.”
“MIT license on both Pro and Flash weights is the real story. Fine-tune, redistribute, ship commercially, no lawyer needed. Plus 384K output tokens for long-doc generation.”
“MIT weights, 1M context, and output tokens roughly 29x cheaper than Opus 4.8. Cache hits make input basically free. Moved my bulk pipelines over and the bill collapsed.”
DeepSeek-V4: frequently asked questions
How much does DeepSeek-V4 cost per 1M tokens?
DeepSeek-V4 costs $1.32/1M in at peak, $0.66 off-peak (cache hit from $0.022) per 1M input tokens and $3.96/1M out at peak, $1.98 off-peak per 1M output tokens. Official DeepSeek API price for deepseek-v4-pro (DeepSeek-V4-Pro-0813): peak $1.32/1M input (cache hit $0.044), $3.96/1M output; off-peak half price ($0.66/$1.98, cache hit $0.022). Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday. V4-Flash is retired: deepseek-v4-flash now routes to V4.1-Flash at Flash prices. Verified against api-docs.deepseek.com 2026-11-10.