DeepSeek
DeepSeek V4 Flash is listed by DeepSeek at $0.15 input and $0.6 output per million tokens. The recorded context window is 1,048,576 tokens. Release date on file: 2026-10-17. LMArena's text board published 2026-10-20 gives it an overall rating of 1432.0 (rank 92, 51,959 votes). Price source: models.dev (MIT).
models.dev · MIT · retrieved Oct 24, 2026
Split quote: LiteLLM lists $0.3 / $1.2, more than 5% off this blended price. Both are shown; neither is overwritten.
Input and output list price. A missing day stays a gap in the line. The series is not a quality history.
LMArena is CC-BY-4.0. Epoch scores, including LiveBench where Epoch publishes it, are CC-BY-4.0. A blank cell means this snapshot has no match. Nothing here is estimated.
| Benchmark | Score | Detail | Source |
|---|---|---|---|
| LMArena overall | 1432.0 | rank 92 · 51,959 votes | LMArena leaderboard dataset · CC-BY-4.0 |
| LMArena coding | 1456.8 | rank 89 · 15,284 votes | LMArena leaderboard dataset · CC-BY-4.0 |
| LMArena math | 1425.9 | rank 110 · 2,699 votes | LMArena leaderboard dataset · CC-BY-4.0 |
| GPQA Diamond | — | — | — |
| LiveBench | — | — | — |
| SWE-bench Verified | — | — | — |
| Epoch Capability Index | 146.11 | 143.82–147.99 | Epoch AI Capabilities & benchmarking · CC-BY-4.0 |
The list row is the price used for the value index. Other rows are the same model id on models.dev, plus the LiteLLM catalog price when we have one. A split quote is kept beside the list price.
| Quote | Provider | Input / 1M | Output / 1M | Blended | Source |
|---|---|---|---|---|---|
| List | DeepSeek | $0.15 | $0.60 | $0.487 | models.dev · MIT |
| Also listed | Standard Compute | $0.083 | $0.17 | $0.145 | models.dev · MIT |
| Also listed | Merge Gateway | $0.30 | $1.20 | $0.975 | models.dev · MIT |
| Also listed | Impossibl | $0.19 | $0.51 | $0.430 | models.dev · MIT |
| Split quote | deepseek | $0.30 | $1.20 | $0.975 | LiteLLM model_prices_and_context_window.json · MIT |
Spec fields are from the models.dev row for DeepSeek.