Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

Start free brief
BenchLM
Advertise

Model comparison

Grok 4.6 vs Kimi K3

Updated August 12, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

Grok 4.6

xAI

—

Evidence status unavailable

90% interval unavailable

Kimi K3

Moonshot AI

80.5/100

Supported · Public rank #5

90% interval 77.3–83.6

At least one model is not scored in the current public ranking lane, so the page does not name an overall quality winner.

2 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

  • Long documents

    Prompts that approach the documented context limit

    Kimi K3

    Kimi K3 has the larger documented context window.

    Confidence: documented

  • Chat turn cost

    1K fresh input + 500 output tokens

    Grok 4.6

    Grok 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Grok 4.6

    Grok 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

Show secondary and unsupported calls

  • Repository review cost

    50K fresh input + 3K output tokens

    Grok 4.6

    Grok 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

    No shared weighted benchmark basis supports a winner.

    Confidence: limited

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    No shared weighted benchmark basis supports a winner.

    Confidence: limited

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

Shared results
2
Grok 4.6 only
3
Kimi K3 only
33
Like-for-like categories
0 / 8

Category results, on a stated basis

Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.

Agentic

Not comparable
Grok 4.6
Not measured
Kimi K3
89.5
Weighted basis
0 vs 2 rows
Reading
Not comparable

Coding

Not comparable
Grok 4.6
Not measured
Kimi K3
Not measured
Weighted basis
0 vs 0 rows
Reading
Not comparable

Reasoning

Not comparable
Grok 4.6
Not measured
Kimi K3
Not measured
Weighted basis
0 vs 0 rows
Reading
Not comparable

Knowledge

Not comparable
Grok 4.6
Not measured
Kimi K3
61.0
Weighted basis
0 vs 2 rows
Reading
Not comparable

Math

Not comparable
Grok 4.6
Not measured
Kimi K3
Not measured
Weighted basis
0 vs 0 rows
Reading
Not comparable

Multilingual

Not comparable
Grok 4.6
Not measured
Kimi K3
Not measured
Weighted basis
0 vs 0 rows
Reading
Not comparable

Multimodal

Not comparable
Grok 4.6
Not measured
Kimi K3
78.5
Weighted basis
0 vs 3 rows
Reading
Not comparable

Instruction following

Not comparable
Grok 4.6
Not measured
Kimi K3
Not measured
Weighted basis
0 vs 0 rows
Reading
Not comparable

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

A shared-evidence shape is not available.

BenchLM does not draw a radar or infer missing axes when the matched evidence is too sparse.

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

Grok 4.6
$0.005
Fits in one request
Kimi K3
$0.0105
Fits in one request

Grok 4.6 has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

Grok 4.6
$0.118
Fits in one request
Kimi K3
$0.195
Fits in one request

Grok 4.6 has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Grok 4.6
$0.2
Fits in one request
Kimi K3
$0.27
Fits in one request

Grok 4.6 has the lower modeled cost

Costs use the listed standard API rates.

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

Grok 4.6

$0.5 per 1M cached input tokens

xAI Grok 4.6 release notes

Kimi K3

$0.3 per 1M cached input tokens

Documented inputs

Grok 4.6

Not sourced

Kimi K3

Not sourced

Documented outputs

Grok 4.6

Not sourced

Kimi K3

Not sourced

Provider availability

Grok 4.6

Not sourced

Kimi K3

Not sourced

Reasoning profile

Grok 4.6

Reasoning

Kimi K3

Reasoning

Weight access

Grok 4.6

Proprietary

Kimi K3

Pending

License

Grok 4.6

Proprietary

Kimi K3

Pending

Release date

Grok 4.6

2026-08-12

Kimi K3

2026-07-16

If you already use one of these models
Deployment change
The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal
At least one model is not scored in the current public ranking lane, so the page does not name an overall quality winner.
Workload cost
Repository review: $0.118 vs $0.195. Cache-heavy agent loop: $0.2 vs $0.27.
Context tradeoff
Kimi K3 has the larger documented window (1.05M).

Run the same representative tasks against both endpoints before changing production traffic.

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence38 rows

Agentic

  • Terminal-Bench 3.0

    Grok 4.626.0%
    Source
    Kimi K3—

    Not directly comparable

  • APEX-Agents

    Grok 4.657.5%
    Source
    Kimi K337.6%
    Source

    Grok 4.6 leads this result

  • Terminal-Bench 2.0

    Grok 4.6—
    Kimi K388.3%
    Source

    Not directly comparable

  • BrowseComp

    Grok 4.6—
    Kimi K391.2%
    Source

    Not directly comparable

  • DeepSearchQA

    Grok 4.6—
    Kimi K395.0%
    Source

    Not directly comparable

  • Toolathlon-Verified

    Grok 4.6—
    Kimi K373.2%
    Source

    Not directly comparable

  • MCP Atlas

    Grok 4.6—
    Kimi K384.2%
    Source

    Not directly comparable

  • AutomationBench

    Grok 4.6—
    Kimi K330.8%
    Source

    Not directly comparable

  • JobBench

    Grok 4.6—
    Kimi K352.9%
    Source

    Not directly comparable

  • SpreadsheetBench 2

    Grok 4.6—
    Kimi K334.8%
    Source

    Not directly comparable

  • DECK-Bench

    Grok 4.6—
    Kimi K373.5%
    Source

    Not directly comparable

Coding

  • deepSwe

    Grok 4.665.9%
    Source
    Kimi K367.5%
    Source

    Kimi K3 leads this result

  • cursorBench32

    Grok 4.669.9%
    Source
    Kimi K3—

    Not directly comparable

  • FrontierCode 1.1 Extended

    Grok 4.661.3%
    Source
    Kimi K3—

    Not directly comparable

  • FrontierSWE

    Grok 4.6—
    Kimi K381.2%
    Source

    Not directly comparable

  • ProgramBench

    Grok 4.6—
    Kimi K377.8%
    Source

    Not directly comparable

  • Kimi Code Bench v2

    Grok 4.6—
    Kimi K372.9%
    Source

    Not directly comparable

  • sweMarathon

    Grok 4.6—
    Kimi K342%
    Source

    Not directly comparable

  • PostTrain Bench

    Grok 4.6—
    Kimi K336.6%
    Source

    Not directly comparable

  • MLS-Bench Lite

    Grok 4.6—
    Kimi K348.3%
    Source

    Not directly comparable

  • VulcanBench v3

    Grok 4.6—
    Kimi K373.9%
    Source

    Not directly comparable

Knowledge

  • GPQA

    Grok 4.6—
    Kimi K393.5%
    Source

    Not directly comparable

  • GPQA-D

    Grok 4.6—
    Kimi K393.5%
    Source

    Not directly comparable

  • HLE

    Grok 4.6—
    Kimi K356%
    Source

    Not directly comparable

  • HLE w/o tools

    Grok 4.6—
    Kimi K343.5%
    Source

    Not directly comparable

Multimodal

  • OfficeQA Pro

    Grok 4.6—
    Kimi K363.3%
    Source

    Not directly comparable

  • MMMU-Pro

    Grok 4.6—
    Kimi K381.6%
    Source

    Not directly comparable

  • MMMU-Pro w/ Python

    Grok 4.6—
    Kimi K383.4%
    Source

    Not directly comparable

  • CharXiv w/o tools

    Grok 4.6—
    Kimi K384.8%
    Source

    Not directly comparable

  • CharXiv

    Grok 4.6—
    Kimi K391.3%
    Source

    Not directly comparable

  • MathVision

    Grok 4.6—
    Kimi K394.3%
    Source

    Not directly comparable

  • MathVision w/ Python

    Grok 4.6—
    Kimi K397.8%
    Source

    Not directly comparable

  • BabyVision w/ Python

    Grok 4.6—
    Kimi K385.7%
    Source

    Not directly comparable

  • ZeroBench

    Grok 4.6—
    Kimi K323.0%
    Source

    Not directly comparable

  • ZeroBench w/ Python

    Grok 4.6—
    Kimi K341.0%
    Source

    Not directly comparable

  • WorldVQA ForceAnswer

    Grok 4.6—
    Kimi K351.0%
    Source

    Not directly comparable

  • OmniDocBench

    Grok 4.6—
    Kimi K391.1%
    Source

    Not directly comparable

  • PerceptionBench

    Grok 4.6—
    Kimi K358.5%
    Source

    Not directly comparable

Frequently asked questions

Which is better, Grok 4.6 or Kimi K3?

At least one model is not scored in the current public ranking lane, so the page does not name an overall quality winner. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, Grok 4.6 or Kimi K3?

The published evidence does not provide a shared weighted coding basis for both models, so BenchLM does not name a coding winner.

Which is better for agentic tasks, Grok 4.6 or Kimi K3?

The published evidence does not provide a shared weighted agentic tasks basis for both models, so BenchLM does not name a agentic tasks winner.

Which costs less, Grok 4.6 or Kimi K3?

For the stated presets, chat costs $0.005 on Grok 4.6 and $0.0105 on Kimi K3; repository review costs $0.118 and $0.195; the cache-heavy agent loop costs $0.2 and $0.27. Costs use the listed standard API rates.

Which has the larger context window, Grok 4.6 or Kimi K3?

Kimi K3 has the larger documented context window: 1.05M, compared with 500K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated August 12, 2026

Watch Grok 4.6 vs Kimi K3

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 2,000+ readers.