Skip to main content
BenchLM
Data

Gemma 4 26B A4B vs Kimi K3

Updated November 17, 2026. Rank says Kimi K3 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInAPI/MCP

Decision reading

Kimi K3 has the higher public point estimate, 67 versus 40.92. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 6 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
Google logo

Google

40.92/100

Estimated · Public rank #136

Conditional range 31.2–50.6

Model B
Moonshot AI logo

Moonshot AI

67/100

Supported · Public rank #20

90% interval 61.6–72.4

API price per 1M tokens

Gemma 4 26B A4B: No comparable API price

Kimi K3: $3 input / $15 output

Context window

Gemma 4 26B A4B: 256K

Kimi K3: 1.05M

Released

Gemma 4 26B A4B: April 2, 2026

Kimi K3: July 16, 2026

Shared results
6
Gemma 4 26B A4B only
4
Kimi K3 only
54
Like-for-like categories
2 / 8
Estimated: Gemma 4 26B A4B · Supported: Kimi K3. Conditional ranges do not establish rank confidence.How the comparison works

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.

  • Coding work

    Code generation, repair, and software-engineering tasks

    Kimi K3

    Kimi K3 has the higher public coding point estimate, 56.3 to 30.8, with Supported evidence for both models. The conditional ranges do not establish rank confidence.

    Confidence: limited
  • Long documents

    Prompts that approach the documented context limit

    Kimi K3

    Kimi K3 has the larger documented context window.

    Confidence: documented
Show secondary and unsupported calls
  • Agentic work

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

    Not enough matched evidence

    Gemma 4 26B A4B is not ranked on the public lane for agentic, so no winner is named for agentic.

    Confidence: limited
  • Chat turn cost

    1K fresh input + 500 output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates
  • Cache-heavy agent loop cost

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

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates
  • Repository review cost

    50K fresh input + 3K output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

30.8Gemma 4 26B A4B56.3Kimi K3

Like-for-like · BenchAlign v5.8

Kimi K3 has the higher coding point estimate. Conditional score ranges do not establish rank confidence.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Same basis rules as the category table below

Or describe your own job

The LLM Selector reads what you need a model for and returns a shortlist, ranked in the same order the job pages publish.

Describe the task and constraints. This selector recommends models; it cannot carry out the task or access your files and accounts.

We keep what you type here, drafts included, for 180 days, and after that only the job and requirements read from it. Nothing is tied to your IP address or account. Leave out names and other personal details. Privacy

What is actually comparable

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

2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

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.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, reasoning, multimodal, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Coding

Like-for-like
Gemma 4 26B A4B
30.8
Supported · #104/163
Kimi K3
56.3
Supported · #23/163
Basis
BenchAlign v5.8 lane · 2 vs 18 public rows
Reading
Kimi K3 leads

Reasoning

Like-for-like
Gemma 4 26B A4B
35.0
Supported · #102/161
Kimi K3
65.4
Supported · #21/161
Basis
BenchAlign v5.8 lane · 1 vs 2 public rows
Reading
Kimi K3 leads

Multimodal

Directional only
Gemma 4 26B A4B
47.7
Supported · #56/67
Kimi K3
66.7
Estimated · #19/67
Basis
BenchAlign v5.8 lane · 3 vs 14 public rows
Reading
Directional only

Knowledge

Directional only
Gemma 4 26B A4B
41.2
Estimated · #116/181
Kimi K3
67.5
Supported · #20/181
Basis
BenchAlign v5.8 lane · 4 vs 6 public rows
Reading
Directional only

Agentic

Not comparable
Gemma 4 26B A4B
Not ranked
Kimi K3
65.1
Supported · #21/140
Basis
BenchAlign v5.8 lane · 0 vs 19 public rows
Reading
Not comparable

Multilingual

Not comparable
Gemma 4 26B A4B
Not ranked
Kimi K3
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Gemma 4 26B A4B
88.7
#37/157
Kimi K3
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Gemma 4 26B A4B
Not ranked
Kimi K3
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, reasoning, multimodal, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Supported evidence per lane · bars run 0–100Methodology

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

Gemma 4 26B A4B
Self-hosted; infrastructure cost varies
Fits in one request
Kimi K3
$0.0105
Fits in one request

Gemma 4 26B A4B has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Gemma 4 26B A4B
Self-hosted; infrastructure cost varies
Fits in one request
Kimi K3
$0.195
Fits in one request

Gemma 4 26B A4B has no comparable published API token rate.

Cache-heavy agent loop

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

Gemma 4 26B A4B
Self-hosted; infrastructure cost varies
Fits in one request
Cached-input rate unavailable
Kimi K3
$0.27
Fits in one request

Gemma 4 26B A4B has no comparable published API token rate.

Cached input falls back to the list input rate only where a cached rate is unpublished

Price against score, among all priced models

Overall score against output price per 1M tokens, one model per family. Pricing as of November 17, 2026.

Kimi K3

20406080100$0.10$0.50$1$5$10$50$100Output price, $ per 1M tokens (log)Overall score
  • Gemma 4 26B A4B has no public API price, so it is not plotted.
  • 3 cheaper models score at least as high as Kimi K3. The nearest is Gemini 4 Carbon (score 85.1, $10.00 per 1M output tokens).

Gray dots are the other priced models. The dark step line is the efficiency frontier: for a model under it, another model costs no more and scores at least as high.

See Kimi K3 on the full price vs performance chart

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.

Gemma 4 26B A4B

No comparable hosted API rate

Kimi K3

$0.3 per 1M cached input tokens

Provider availability

Gemma 4 26B A4B

Generally Available · Gemini API, Google AI Studio, open weights

Google Gemma Gemini API guide

Kimi K3

Not sourced

Reasoning profile

Gemma 4 26B A4B

Reasoning

Kimi K3

Reasoning

Weight access

Gemma 4 26B A4B

Open Weight

Kimi K3

Open Weight

License

Gemma 4 26B A4B

Open Weight

Kimi K3

Open Weight

Release date

Gemma 4 26B A4B

2026-04-02

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
Kimi K3 has the higher public point estimate, 67 versus 40.92. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
Kimi K3 has the larger documented window (1.05M).
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Gemma 4 26B A4B or Kimi K3?

Kimi K3 has the higher public point estimate, 67 versus 40.92. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, Gemma 4 26B A4B or Kimi K3?

Kimi K3 has the higher public coding point estimate, 56.3 to 30.8, with Supported evidence for both models. The conditional ranges do not establish rank confidence.

Which is better for agentic tasks, Gemma 4 26B A4B or Kimi K3?

Gemma 4 26B A4B is not ranked on the public lane for agentic tasks, so no winner is named for agentic tasks.

Which costs less, Gemma 4 26B A4B or Kimi K3?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, Gemma 4 26B A4B or Kimi K3?

Kimi K3 has the larger documented context window: 1.05M, compared with 256K.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

Gemma 4 26B A4B
API / mo (Novita AI)$398
Self-host / mo$298
Break-even37M/day
Kimi K3
API / mo$13,500
Self-host / mo$57,606
Break-evenNot reached at 70% load
Model the full break-even

Benchmark evidence

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

Browse raw public benchmark evidence64 rows

Agentic

  • Terminal-Bench 4.1

    Gemma 4 26B A4B—
    Kimi K323.00%
    Source

    Not directly comparable

  • Finch

    Gemma 4 26B A4B—
    Kimi K377.3%
    Source

    Not directly comparable

  • b3 Agent Security

    Gemma 4 26B A4B—
    Kimi K388.0%
    Source

    Not directly comparable

  • Cyber Misuse Refusal

    Gemma 4 26B A4B—
    Kimi K394.0%
    Source

    Not directly comparable

  • Terminal-Bench 2.1

    Gemma 4 26B A4B—
    Kimi K388.3%
    Source

    Not directly comparable

  • BrowseComp

    Gemma 4 26B A4B—
    Kimi K391.2%
    Source

    Not directly comparable

  • DeepSearchQA

    Gemma 4 26B A4B—
    Kimi K395.0%
    Source

    Not directly comparable

  • Toolathlon-Verified

    Gemma 4 26B A4B—
    Kimi K373.2%
    Source

    Not directly comparable

  • MCP Atlas

    Gemma 4 26B A4B—
    Kimi K384.2%
    Source

    Not directly comparable

  • AutomationBench

    Gemma 4 26B A4B—
    Kimi K330.8%
    Source

    Not directly comparable

  • JobBench

    Gemma 4 26B A4B—
    Kimi K352.9%
    Source

    Not directly comparable

  • APEX-Agents

    Gemma 4 26B A4B—
    Kimi K337.6%
    Source

    Not directly comparable

  • SpreadsheetBench 2

    Gemma 4 26B A4B—
    Kimi K334.8%
    Source

    Not directly comparable

  • DECK-Bench

    Gemma 4 26B A4B—
    Kimi K373.5%
    Source

    Not directly comparable

  • Terminal-Bench 2.1 (Vals)

    Gemma 4 26B A4B—
    Kimi K380.9%
    Source

    Not directly comparable

  • ApprenticeBench

    Gemma 4 26B A4B—
    Kimi K318%
    Source

    Not directly comparable

  • Hermes Index

    Gemma 4 26B A4B—
    Kimi K339.2%
    Source

    Not directly comparable

  • Hermes Bench

    Gemma 4 26B A4B—
    Kimi K371.5%
    Source

    Not directly comparable

  • MazeBench

    Gemma 4 26B A4B—
    Kimi K31%
    Source

    Not directly comparable

Coding

  • ScarfBench

    Shared source
    Gemma 4 26B A4B24.4%
    Kimi K361.3%

    Kimi K3 leads this result

  • LiveCodeBench v6

    Gemma 4 26B A4B77.1%
    Source
    Kimi K3—

    Not directly comparable

  • DeepSWE

    Gemma 4 26B A4B—
    Kimi K367.5%
    Source

    Not directly comparable

  • CursorBench 3.2

    Gemma 4 26B A4B—
    Kimi K360.8%
    Source

    Not directly comparable

  • FrontierSWE

    Gemma 4 26B A4B—
    Kimi K381.2%
    Source

    Not directly comparable

  • ProgramBench

    Gemma 4 26B A4B—
    Kimi K377.8%
    Source

    Not directly comparable

  • Kimi Code Bench v2

    Gemma 4 26B A4B—
    Kimi K372.9%
    Source

    Not directly comparable

  • sweMarathon

    Gemma 4 26B A4B—
    Kimi K342%
    Source

    Not directly comparable

  • PostTrainBench (lab runs)

    Gemma 4 26B A4B—
    Kimi K336.6%
    Source

    Not directly comparable

  • MLS-Bench Lite

    Gemma 4 26B A4B—
    Kimi K348.3%
    Source

    Not directly comparable

  • VulcanBench v3

    Gemma 4 26B A4B—
    Kimi K373.7%
    Source

    Not directly comparable

  • OpenHarmony Bench v1.0

    Gemma 4 26B A4B—
    Kimi K357.3%
    Source

    Not directly comparable

  • FrontierSWE v2

    Gemma 4 26B A4B—
    Kimi K325.9%
    Source

    Not directly comparable

  • LiveCodeBench (Vals)

    Gemma 4 26B A4B—
    Kimi K387.2%
    Source

    Not directly comparable

  • SWE-bench (Vals)

    Gemma 4 26B A4B—
    Kimi K393.4%
    Source

    Not directly comparable

  • PostTrainBench v1.1

    Gemma 4 26B A4B—
    Kimi K332.0%
    Source

    Not directly comparable

  • Bug Hunt Bench

    Gemma 4 26B A4B—
    Kimi K321.0 fixes
    Source

    Not directly comparable

  • FrontierCode 2.0

    Gemma 4 26B A4B—
    Kimi K318.9%
    Source

    Not directly comparable

  • PostTrainBench

    Gemma 4 26B A4B—
    Kimi K331.5%
    Source

    Not directly comparable

Reasoning

  • MRCRv2

    Gemma 4 26B A4B44.1%
    Source
    Kimi K3—

    Not directly comparable

  • ARC-AGI-1

    Gemma 4 26B A4B—
    Kimi K394.50%
    Source

    Not directly comparable

  • ARC-AGI-2

    Gemma 4 26B A4B—
    Kimi K360.4%
    Source

    Not directly comparable

Multimodal

  • ExtractBench (LlamaIndex)

    Shared source
    Gemma 4 26B A4B70.27%
    Kimi K383.17%

    Kimi K3 leads this result

  • ParseBench-1

    Gemma 4 26B A4B58.50%
    Source
    Kimi K3—

    Not directly comparable

  • MMMU-Pro

    Gemma 4 26B A4B73.8%
    Source
    Kimi K381.6%
    Source

    Kimi K3 leads this result

  • OfficeQA Pro

    Gemma 4 26B A4B—
    Kimi K363.3%
    Source

    Not directly comparable

  • MMMU-Pro w/ Python

    Gemma 4 26B A4B—
    Kimi K383.4%
    Source

    Not directly comparable

  • CharXiv w/o tools

    Gemma 4 26B A4B—
    Kimi K384.8%
    Source

    Not directly comparable

  • CharXiv

    Gemma 4 26B A4B—
    Kimi K391.3%
    Source

    Not directly comparable

  • MathVision

    Gemma 4 26B A4B—
    Kimi K394.3%
    Source

    Not directly comparable

  • MathVision w/ Python

    Gemma 4 26B A4B—
    Kimi K397.8%
    Source

    Not directly comparable

  • BabyVision w/ Python

    Gemma 4 26B A4B—
    Kimi K385.7%
    Source

    Not directly comparable

  • ZeroBench

    Gemma 4 26B A4B—
    Kimi K323.0%
    Source

    Not directly comparable

  • ZeroBench w/ Python

    Gemma 4 26B A4B—
    Kimi K341.0%
    Source

    Not directly comparable

  • WorldVQA ForceAnswer

    Gemma 4 26B A4B—
    Kimi K351.0%
    Source

    Not directly comparable

  • OmniDocBench

    Gemma 4 26B A4B—
    Kimi K391.1%
    Source

    Not directly comparable

  • PerceptionBench

    Gemma 4 26B A4B—
    Kimi K358.5%
    Source

    Not directly comparable

Knowledge

  • GPQA

    Gemma 4 26B A4B82.3%
    Source
    Kimi K393.5%
    Source

    Kimi K3 leads this result

  • MMLU-Pro

    Gemma 4 26B A4B82.6%
    Source
    Kimi K3—

    Not directly comparable

  • HLE

    Gemma 4 26B A4B17.2%
    Source
    Kimi K356%
    Source

    Kimi K3 leads this result

  • HLE w/o tools

    Gemma 4 26B A4B8.7%
    Source
    Kimi K343.5%
    Source

    Kimi K3 leads this result

  • GPQA-D

    Gemma 4 26B A4B—
    Kimi K393.5%
    Source

    Not directly comparable

  • GPQA Diamond (Vals)

    Gemma 4 26B A4B—
    Kimi K392.9%
    Source

    Not directly comparable

  • MMLU-Pro (Vals)

    Gemma 4 26B A4B—
    Kimi K388.0%
    Source

    Not directly comparable

Instruction following

  • Gray Swan IPI (15 attempts)

    Gemma 4 26B A4B—
    Kimi K352.7%
    Source

    Not directly comparable

64 public results · 6 shared

Watch Gemma 4 26B A4B vs Kimi K3

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

Read a sample issue

Join 7,000+ subscribers.

Last updated November 17, 2026