DeepSeek V4 Pro vs Kimi K3

As of October 21, 2026, Kimi K3 costs 127% more per input token than DeepSeek V4 Pro ($3.00 vs $1.32 per 1M) and has a 1,048,576-token context window versus 1,000,000.

Prices from LiteLLM's price fileChecked October 21, 2026

DeepSeek V4 Pro vs Kimi K3 side by side

Headline prices in US dollars per 1M tokens: each model's own lab where it sells it, otherwise its cheapest provider. Source: LiteLLM model price file, as of October 21, 2026.
	DeepSeek V4 Pro	Kimi K3
Input, per 1M tokens	$1.32	$3.00
Output, per 1M tokens	$3.96	$15.00
Cache read, per 1M tokens	$0.044	$0.30
Cache write, per 1M tokens	$0	–
Context window	1,000,000 tokens	1,048,576 tokens
Max output	393,216 tokens	1,048,576 tokens
Image input	No	Yes
Tool calling	Yes	Yes
Structured output (JSON)	Yes	Yes
Prompt caching	Yes	Yes
Reasoning	Yes	Yes
Providers on LiteLLM	18	15
Price shown is from	DeepSeek	Moonshot AI
Added to LiteLLM	June 2026	August 16, 2026

Cheapest provider for each

DeepSeek V4 Pro: Sail at $0.92 input and $2.77 output per 1M tokens (sail/deepseek-ai/DeepSeek-V4-Pro-0813)
Kimi K3: Sail at $2.50 input and $12.50 output per 1M tokens (sail/moonshotai/Kimi-K3)

Benchmark scores

From the Epoch AI Benchmarking Hub (CC BY 4.0), data as of August 27, 2026. Some results were run by Epoch and others collected by Epoch from each benchmark's own leaderboard, linked in the first column. Each model's best reported reasoning setting is shown in brackets. Higher is better. Only benchmarks with a score for both models are shown.
Benchmark	DeepSeek V4 Pro	Kimi K3
Coding and agents
FrontierCode	17.6% (none)	44.2%
APEX-Agents	47.3%	47.8%
Reasoning and knowledge
GPQA Diamond	90.9% (high)	93.1% (max)
ARC-AGI-2	61.3% (max)	60.4% (max)
SimpleQA Verified	47.0% (max)	50.6% (max)
Math
FrontierMath (tiers 1-3)	45.3% (max)	72.2% (max)
FrontierMath (tier 4)	2.4% (max)	39.0% (max)
OTIS Mock AIME 2024-2025	96.7% (max)	97.2% (max)

When to pick which

Pick DeepSeek V4 Pro for

Costs 56% less per input token: $1.32 vs $3.00 per 1M.

Pick Kimi K3 for

Supports image input, which DeepSeek V4 Pro doesn't list.
Returns up to 1,048,576 tokens per response, vs 393,216.

From differences in the data above only; where the data is equal, no reason is given.

Call DeepSeek V4 Pro or Kimi K3 with LiteLLM

Both use the OpenAI format through LiteLLM, so switching is a model name change. Set each provider's API key in the environment first.

Python SDK

from litellm import completion

messages = [{"role": "user", "content": "Hello!"}]

# DeepSeek V4 Pro
response = completion(model="deepseek-v4-pro", messages=messages)

# Kimi K3
response = completion(model="moonshot/kimi-k3", messages=messages)

# DeepSeek V4 Pro, falling back to Kimi K3 if the call fails
response = completion(model="deepseek-v4-pro", messages=messages, fallbacks=["moonshot/kimi-k3"])

FAQ

Which is cheaper, DeepSeek V4 Pro or Kimi K3?

DeepSeek V4 Pro is cheaper at 3 input tokens per output token: $1.98 per 1M tokens blended ($1.32 input, $3.96 output), versus $6.00 ($3.00 and $15.00) for Kimi K3, as of October 21, 2026.

Which has the larger context window, DeepSeek V4 Pro or Kimi K3?

Kimi K3 has the larger context window: 1,048,576 tokens, versus 1,000,000 for DeepSeek V4 Pro.

Can I switch from DeepSeek V4 Pro to Kimi K3 without changing my code?

Yes. Through LiteLLM both use the OpenAI format, so switching means changing the model name from "deepseek-v4-pro" to "moonshot/kimi-k3", or pointing a LiteLLM gateway model alias at the other model. You can also use one as a fallback for the other.