Sign Up

Decision support

Gemini 4 Carbon vs Kimi K3.1 Pro

Compare canonical objects directly. The useful dimensions change with the object type, so model and hardware comparisons stay separate.

ModelsHardware
Gemini 4 Carbon

Google DeepMind

Sources:models.devOpenRouter

Kimi K3.1 Pro

Moonshot AI

Sources:models.devOpenRouter

Benchmark results

Head to head on 3 benchmarks both models report

  • Gemini 4 Carbon3leads
  • Kimi K3.1 Pro0leads

  • Arena Text

    Arena rating · Higher is better

    • Gemini 4 Carbon
      Best1526
      +43 ahead · Arena #1CI 1516 to 1536
    • Kimi K3.1 Pro
      1483
      Arena #24CI 1472 to 1494
    Runs and sources
    Gemini 4 Carbon
    • high152695% CI 1516 to 1536 · 4,282 votes
    Kimi K3.1 Pro
    • max148395% CI 1472 to 1494 · 3,443 votes
    Source:Arena Text leaderboard, 2026-11-09

  • Arena WebDev

    Arena rating · Higher is better

    • Gemini 4 Carbon
      Best1768
      +80 ahead · Arena #4CI 1753 to 1783
    • Kimi K3.1 Pro
      1688
      Arena #11CI 1678 to 1697
    Runs and sources
    Gemini 4 Carbon
    • high176895% CI 1753 to 1783 · 2,120 votes
    Kimi K3.1 Pro
    • max168895% CI 1678 to 1697 · 4,916 votes
    Source:Arena WebDev leaderboard, 2026-11-10

  • Agent Arena

    IPS score · Higher is better

    • Gemini 4 Carbon
      Best+0.170
      +0.121 ahead · Arena #3CI +0.163 to +0.177
    • Kimi K3.1 Pro
      +0.049
      Arena #15CI +0.038 to +0.060
    Runs and sources
    Gemini 4 Carbon
    • high+0.17095% CI +0.163 to +0.177 · 9,558,978 observations
    Kimi K3.1 Pro
    • max+0.04995% CI +0.038 to +0.060 · 6,446,421 observations
    Source:Agent Arena leaderboard, 2026-11-11

  • ARC-AGI-2

    Best published score · Higher is better

    • Gemini 4 Carbon
      94.2%
      #2 of 85
    • Kimi K3.1 Pro
      Not reported
    Runs and sources
    Gemini 4 Carbon
    • low49.6%
    • medium94.2%
    • high92.1%
    Kimi K3.1 Pro

    Not reported

    Source:External evaluation via Epoch AIOriginal-source terms apply

  • FrontierMath-Tier-4-v2-Private

    Best published score · Higher is better

    • Gemini 4 Carbon
      95.1%
      #4 of 72
    • Kimi K3.1 Pro
      Not reported
    Runs and sources
    Gemini 4 Carbon
    • high95.1%
    Kimi K3.1 Pro

    Not reported

    Source:Epoch AI evaluation

  • FrontierMath-Tiers-1-3-v2-Private

    Best published score · Higher is better

    • Gemini 4 Carbon
      90.5%
      #6 of 87
    • Kimi K3.1 Pro
      Not reported
    Runs and sources
    Gemini 4 Carbon
    • high90.5%
    Kimi K3.1 Pro

    Not reported

    Source:Epoch AI evaluation

  • GPQA diamond

    Best published score · Higher is better

    • Gemini 4 Carbon
      96.5%
      #1 of 161
    • Kimi K3.1 Pro
      Not reported
    Runs and sources
    Gemini 4 Carbon
    • high96.5%
    Kimi K3.1 Pro

    Not reported

    Source:Epoch AI evaluation

  • OTIS Mock AIME 2024-2025

    Best published score · Higher is better

    • Gemini 4 Carbon
      100.0%
      #1 of 158
    • Kimi K3.1 Pro
      Not reported
    Runs and sources
    Gemini 4 Carbon
    • high100.0%
    Kimi K3.1 Pro

    Not reported

    Source:Epoch AI evaluation

  • SimpleQA Verified

    Best published score · Higher is better

    • Gemini 4 Carbon
      77.7%
      #2 of 87
    • Kimi K3.1 Pro
      Not reported
    Runs and sources
    Gemini 4 Carbon
    • high77.7%
    Kimi K3.1 Pro

    Not reported

    Source:Epoch AI evaluation

Each model's best published result is shown; "Not reported" means no matched evaluation, not a score of zero. Arena and DeepSWE ranks are the leaderboards' own; other ranks are among catalogue models with a score. Results from different harnesses or settings are not directly comparable.

Original-source terms apply to evaluations credited to their original publishers.

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Arena ratings: Arena (lmarena-ai/leaderboard-dataset), licensed CC BY 4.0. Ratings come from pairwise human votes and are only comparable within one leaderboard; the interval is Arena's 95% confidence interval. Model names are matched to the Vertical Tabs catalog.

Model specifications

Release date
October 28, 2026
October 23, 2026
Context window
1M
1M
Native context
—
1M
Max output tokens
943,718
943,718
Knowledge cutoff
2026-03
—
Open weights
Closed
Open weights
Parameter count
—
2.8T

API pricing · USD per 1M tokens

Input
$2.00
$0.2375
Output
$10.00
$8.7856
Cached input
$0.10
$0.19
Cache write
$0.1875
—

Capabilities and modalities

Input modalities
audio, file, image, text, video
text, image, video
Output modalities
text
text
Tool calling
Supported
Supported
Reasoning option
Supported
Supported
Supported options
include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_p
frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p
Capabilities
—
vision, thinking, completion, tools

Architecture and availability

Architecture
—
kimi-k3
Model families
—
kimi-k3
Tokenizer
Gemini
Other
Instruction format
—
—
License
Proprietary
—
Providers
2
17
Provider endpoints
14
46
Ollama variants
0
0
Metadata source
ai-research
ai-research

A dash means the field is not reported. Provider-specific performance and pricing may differ.

To give you the best experience, we use cookies for personalization, insights, and performance. Learn more.