MDL
DeepSeek V4 Pro
DeepSeek
Verified against Artificial Analysis · Jul 13, 2026
Capability
94
percentile index
Price /M
$0.435 / $0.87
in / out
Context
1M
tokens
Avg score
65.7
7 benchmarks
AA Intelligence
44.3
index
Speed
64 tok/s
TTFT 1.1s
Released
Apr 24, 2026
About
DeepSeek V4 Pro is a open-weights model in the DeepSeek family from DeepSeek. Benchmarked on 7 evals, averaging 65.7.
Benchmark scores · 7
avg 65.7 — higher is betterThe research behind this model
all papers mentioning it →Scaling the Horizon, Not the Parameters: Reaching Trillion-Parameter Performance with a 35B Agent▲ 98 on HF · Jun 29, 2026FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention▲ 61 on HF · Jun 8, 2026K-BrowseComp: A Web Browsing Agent Benchmark Grounded in Korean Contexts▲ 55 on HF · Jun 1, 2026DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation▲ 31 on HF · Jul 6, 2026AutoTrainess: Teaching Language Models to Improve Language Models Autonomously▲ 23 on HF · Jun 30, 2026
Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.
Performance & price
Output speed64 tok/s
Latency (TTFT)1.1s
First answer1m 9s
Price in / out$0.435 / $0.87 per 1M
Blended price$0.544 / 1M
Coding index59.4
Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →
History
full ledger →Capability over time
Append-only ledger — every observed change to this model's numbers.
API providers
Artificial Analysis$0.87/M