// head_to_head
DeepSeek V4 Pro 0423 vs Kimi K2 Thinking
Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.
Kimi K2 Thinking is the cheaper of the two; neither can be ranked on quality here.
Kimi K2 Thinking does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
deepseek
DeepSeek V4 Pro 0423
- Blended / 1M
- $1.19
- Context
- 1.0M
- Released
- Apr 24, 2026
- Overall score
- 71.6
moonshotai
Kimi K2 Thinking
- Blended / 1M
- $1.07
- Context
- 262K
- Released
- Nov 6, 2025
- Overall score
- Not evaluated
Specs and pricing
| Metric | DeepSeek V4 Pro 0423 | Kimi K2 Thinking |
|---|---|---|
| LiveBench overall Mean of the seven LiveBench category scores, 0–100. Higher is better. | 71.6 | — |
| Cost per point Measured benchmark spend divided by overall score — dollars per point of capability. | $0.0261 | — |
| Blended price / 1M 3:1 input:output mix, the usual shape of production traffic. | $1.19 | $1.07win |
| Input price / 1M | $0.955 | $0.600win |
| Output price / 1M | $1.91win | $2.50 |
| Cached input / 1M Price of an input token served from the prompt cache, where the provider publishes one. | $0.080win | $0.150 |
| Context window | 1.0Mwin | 262K |
| Max output tokens | 384Kwin | 98K |
Benchmarks by category
An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — DeepSeek V4 Pro 0423 on top, Kimi K2 Thinking below, both out of 100.
What each one costs to run
Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.
| Workload | DeepSeek V4 Pro 0423 | Kimi K2 Thinking |
|---|---|---|
| Support chatbot 1.2K in / 400 out × 200K requests | $319.06/mo | $311.60/mo |
| RAG assistant 8K in / 600 out × 100K requests | $528.58/mo | $450.00/mo |
| Coding agent 40K in / 4K out × 20K requests | $426.68/mo | $428.00/mo |
| Document extraction 20K in / 1.5K out × 50K requests | $1,055/mo | $765.00/mo |
| Bulk classification 500 in / 20 out × 5M requests | $2,141/mo | $1,525/mo |
Which should you pick?
You need to fit large documents in one call
DeepSeek V4 Pro 0423
Wider context window — 1.0M against 262K.
You are cost-constrained
Kimi K2 Thinking
Cheaper on blended list price at $1.07 per million tokens.
DeepSeek V4 Pro 0423 vs Kimi K2 Thinking FAQ
Which is better, DeepSeek V4 Pro 0423 or Kimi K2 Thinking?
Kimi K2 Thinking is the cheaper of the two; neither can be ranked on quality here. Kimi K2 Thinking does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
Is DeepSeek V4 Pro 0423 cheaper than Kimi K2 Thinking?
Kimi K2 Thinking is cheaper. On a 3:1 input:output blend, DeepSeek V4 Pro 0423 lists at $1.19 per million tokens and Kimi K2 Thinking at $1.07 — Kimi K2 Thinking is 11% cheaper. Input and output are priced separately — DeepSeek V4 Pro 0423 charges $0.955 in and $1.91 out, Kimi K2 Thinking charges $0.600 and $2.50 — so the model that looks cheaper flips depending on how output-heavy your workload is.
Does DeepSeek V4 Pro 0423 or Kimi K2 Thinking have a bigger context window?
DeepSeek V4 Pro 0423 has the larger context window: 1.0M for DeepSeek V4 Pro 0423 against 262K for Kimi K2 Thinking. Note that a window you can fill is not a window you should fill — retrieval quality usually degrades well before the limit, and you pay for every token you put in it.
Do DeepSeek V4 Pro 0423 and Kimi K2 Thinking support prompt caching?
Both publish a cached-input rate: $0.080 per million for DeepSeek V4 Pro 0423 and $0.150 for Kimi K2 Thinking, against full input rates of $0.955 and $0.600. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.
Related comparisons
How these numbers are produced
- Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
- Scores — LiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
- Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
- “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.
Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.