Grok-3 vs Kimi K2-Thinking-0905
Grok-3 and Kimi K2-Thinking-0905 are closely matched at 30.4 and 36.0 on the LLM Stats Score. Kimi K2-Thinking-0905 is 7.0x cheaper per token.
xAI · Moonshot AI · Updated for 2026
Which is better?
Grok-3 and Kimi K2-Thinking-0905 are closely matched on the overall LLM Stats Score at 30.4 and 36.0.
The models split the 2 individual benchmarks reported for both models evenly.
On price, Kimi K2-Thinking-0905 is roughly 7.0x cheaper per token on a blended 3:1 input/output basis, which adds up quickly at production volume.
Kimi K2-Thinking-0905 also accepts a larger context window (262,144 input tokens), making it the stronger choice for long documents and large codebases.
Based on current LLM Stats indexes, shared benchmarks, pricing, and model metadata for 2026.
Choose Grok-3
- you want predictable pricing at $3.00/M input and $15.00/M output
Choose Kimi K2-Thinking-0905
- cost matters — it's about 7.0x cheaper per token
- you process long inputs — it offers a 262,144 token context window
- you want the most recent training data — it shipped Sep 2025
- you need open weights you can self-host or fine-tune
At a glance
The differences that matter most.
Capability indexes
Additional strengths measured across groups of related public benchmarks
Individual benchmarks
5 reported for Grok-3 · 21 for Kimi K2-Thinking-0905
Grok-3 outperforms in 1 benchmarks (GPQA), while Kimi K2-Thinking-0905 is better at 1 benchmark (AIME 2025).
Both models are evenly matched across the benchmarks.
Human preference
Blind head-to-head votes and playground preference scores
Pricing Analysis
Price comparison per million tokens
For input processing, Grok-3 ($3.00/1M tokens) is 6.4x more expensive than Kimi K2-Thinking-0905 ($0.47/1M tokens).
For output processing, Grok-3 ($15.00/1M tokens) is 7.5x more expensive than Kimi K2-Thinking-0905 ($2.00/1M tokens).
In conclusion, Grok-3 is more expensive than Kimi K2-Thinking-0905.*
* Using a 3:1 ratio of input to output tokens
Context Window
Maximum input and output token capacity
Kimi K2-Thinking-0905 accepts 262,144 input tokens compared to Grok-3's 128,000 tokens. Kimi K2-Thinking-0905 can generate longer responses up to 262,144 tokens, while Grok-3 is limited to 8,000 tokens.
Input capabilities
Documented input modalities across available providers
Grok-3 supports multimodal inputs, whereas Kimi K2-Thinking-0905 does not.
Grok-3 can handle both text and other forms of data like images, making it suitable for multimodal applications.
Grok-3
Kimi K2-Thinking-0905
License
Usage and distribution terms
Grok-3 is licensed under a proprietary license, while Kimi K2-Thinking-0905 uses MIT.
License differences may affect how you can use these models in commercial or open-source projects.
Proprietary
Closed source
MIT
Open weights
Release Timeline
When each model was launched
Grok-3 was released on 2025-02-17, while Kimi K2-Thinking-0905 was released on 2025-09-05.
Kimi K2-Thinking-0905 is 7 months newer than Grok-3.
Feb 17, 2025
1.6 years ago
Sep 5, 2025
1.0 years ago
6mo newerKnowledge Cutoff
When training data ends
Grok-3 has a documented knowledge cutoff of 2024-11-17, while Kimi K2-Thinking-0905's cutoff date is not specified.
We can confirm Grok-3's training data extends to 2024-11-17, but cannot make a direct comparison without Kimi K2-Thinking-0905's cutoff date.
Nov 2024
—
Provider Availability
Grok-3 is available from xAI. Kimi K2-Thinking-0905 is available from DeepInfra, Novita, Fireworks.
Grok-3
Kimi K2-Thinking-0905
Outputs Comparison
Judge for yourself.
Run your own prompts against Grok-3 and Kimi K2-Thinking-0905 side-by-side, then vote on the output you prefer.
FAQ
Common questions about Grok-3 vs Kimi K2-Thinking-0905.