Skip to main content
BenchLM

Gemma 4 31B vs Grok 4.6

Updated September 25, 2026. Rank says Grok 4.6 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInSocial cardCSVJSON

Decision reading

The public evidence has no benchmark result shared by both models, so it does not support a quality verdict. Use the documented cost, context, and runtime rows instead. 0 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

44.84/100

Estimated · Public rank #97

90% interval 24.3–65.3

Model B
xAI logo

xAI

69.19/100

Supported · Public rank #15

90% interval 66.3–72.1

Shared results
0
Gemma 4 31B only
8
Grok 4.6 only
17
Like-for-like categories
2 / 8
Estimated: Gemma 4 31B · Supported: Grok 4.6How 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

    Grok 4.6

    Grok 4.6 leads on the public coding lane, 62 to 35.7, with Supported evidence for both models and non-overlapping 90% intervals.

    Confidence: stronger
  • Long documents

    Prompts that approach the documented context limit

    Grok 4.6

    Grok 4.6 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 31B is scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.

    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.

35.7Gemma 4 31B62.0Grok 4.6

Like-for-like · BenchAlign v5.7

Grok 4.6 leads the like-for-like coding row.

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

Same basis rules as the category table below

What is actually comparable

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

1 category rests 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.

A shared-evidence shape is not available.

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

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.7 lane for agentic, coding, 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 31B
35.7
Supported · #72/135
Grok 4.6
62.0
Supported · #14/135
Basis
BenchAlign v5.7 lane · 2 vs 8 public rows
Reading
Grok 4.6 leads

Knowledge

Like-for-like
Gemma 4 31B
40.0
Supported · #91/158
Grok 4.6
69.0
Supported · #13/158
Basis
BenchAlign v5.7 lane · 4 vs 2 public rows
Reading
Grok 4.6 leads

Agentic

Directional only
Gemma 4 31B
23.9
Estimated · #83/105
Grok 4.6
67.9
Supported · #8/105
Basis
BenchAlign v5.7 lane · 1 vs 4 public rows
Reading
Directional only

Reasoning

Not comparable
Gemma 4 31B
70.0
Unranked · 2 rankable rows
Grok 4.6
57.6
#16/19
Basis
Provisional lane · 0 vs 2 weighted rows
Reading
Not comparable

Multimodal

Not comparable
Gemma 4 31B
58.5
#32/50
Grok 4.6
Not ranked
Basis
Provisional lane · 1 vs 0 weighted rows
Reading
Not comparable

Multilingual

Not comparable
Gemma 4 31B
Not ranked
Grok 4.6
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Gemma 4 31B
91.5
#12/124
Grok 4.6
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Gemma 4 31B
Not ranked
Grok 4.6
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, and knowledge denominators (BenchAlign v5.7) 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 31B
Self-hosted; infrastructure cost varies
Fits in one request
Grok 4.6
$0.005
Fits in one request

Gemma 4 31B has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Gemma 4 31B
Self-hosted; infrastructure cost varies
Fits in one request
Grok 4.6
$0.118
Fits in one request

Gemma 4 31B has no comparable published API token rate.

Cache-heavy agent loop

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

Gemma 4 31B
Self-hosted; infrastructure cost varies
Fits in one request
Cached-input rate unavailable
Grok 4.6
$0.2
Fits in one request

Gemma 4 31B has no comparable published API token rate.

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

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 31B

No comparable hosted API rate

Grok 4.6

$0.5 per 1M cached input tokens

xAI Grok 4.6 release notes

Provider availability

Gemma 4 31B

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

Google Gemma Gemini API guide

Grok 4.6

Not sourced

Reasoning profile

Gemma 4 31B

Reasoning

Grok 4.6

Reasoning

Weight access

Gemma 4 31B

Open Weight

Grok 4.6

Proprietary

License

Gemma 4 31B

Open Weight

Grok 4.6

Proprietary

Release date

Gemma 4 31B

2026-04-02

Grok 4.6

2026-08-12

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
The public evidence has no benchmark result shared by both models, so it does not support a quality verdict.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
Grok 4.6 has the larger documented window (500K).
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Gemma 4 31B or Grok 4.6?

The public evidence has no benchmark result shared by both models, so it does not support a quality verdict. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, Gemma 4 31B or Grok 4.6?

Grok 4.6 leads the public coding lane, 62 to 35.7, with Supported evidence for both models and non-overlapping 90% intervals.

Which is better for agentic tasks, Gemma 4 31B or Grok 4.6?

Grok 4.6 scores higher for agentic tasks on the public lane, 67.9 to 23.9. Gemma 4 31B is scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which costs less, Gemma 4 31B or Grok 4.6?

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 31B or Grok 4.6?

Grok 4.6 has the larger documented context window: 500K, compared with 256K.

Self-host vs API cost

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

Gemma 4 31B
API / mo$0
Self-host / mo$429
Break-even—
Grok 4.6
API / mo$6,000
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
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 evidence25 rows

Agentic

  • Gert Labs

    Gemma 4 31B35.26%
    Source
    Grok 4.6—

    Not directly comparable

  • Terminal-Bench 3.0

    Gemma 4 31B—
    Grok 4.626.5%
    Source

    Not directly comparable

  • APEX-Agents

    Gemma 4 31B—
    Grok 4.657.5%
    Source

    Not directly comparable

  • Terminal-Bench 2.1 (Vals)

    Gemma 4 31B—
    Grok 4.678.3%
    Source

    Not directly comparable

  • ApprenticeBench

    Gemma 4 31B—
    Grok 4.613%
    Source

    Not directly comparable

Coding

  • SWE-Rebench

    Gemma 4 31B41.6%
    Source
    Grok 4.6—

    Not directly comparable

  • React Native Evals

    Gemma 4 31B75.2%
    Source
    Grok 4.6—

    Not directly comparable

  • Bug Hunt Bench

    Gemma 4 31B—
    Grok 4.627 fixes
    Source

    Not directly comparable

  • DeepSWE

    Gemma 4 31B—
    Grok 4.665.9%
    Source

    Not directly comparable

  • cursorBench32

    Gemma 4 31B—
    Grok 4.670.8%
    Source

    Not directly comparable

  • FrontierCode 1.1 Extended

    Gemma 4 31B—
    Grok 4.661.3%
    Source

    Not directly comparable

  • VulcanBench v3

    Gemma 4 31B—
    Grok 4.687.0%
    Source

    Not directly comparable

  • FrontierSWE v2

    Gemma 4 31B—
    Grok 4.625.3%
    Source

    Not directly comparable

  • LiveCodeBench (Vals)

    Gemma 4 31B—
    Grok 4.688.2%
    Source

    Not directly comparable

  • SWE-bench (Vals)

    Gemma 4 31B—
    Grok 4.695.6%
    Source

    Not directly comparable

Reasoning

  • ARC-AGI-1

    Gemma 4 31B—
    Grok 4.687.00%
    Source

    Not directly comparable

  • ARC-AGI-2

    Gemma 4 31B—
    Grok 4.667.1%
    Source

    Not directly comparable

  • ARC-AGI-3

    Gemma 4 31B—
    Grok 4.62.1%
    Source

    Not directly comparable

Multimodal

  • MMMU-Pro

    Gemma 4 31B76.9%
    Source
    Grok 4.6—

    Not directly comparable

Knowledge

  • GPQA

    Gemma 4 31B84.3%
    Source
    Grok 4.6—

    Not directly comparable

  • MMLU-Pro

    Gemma 4 31B85.2%
    Source
    Grok 4.6—

    Not directly comparable

  • HLE

    Gemma 4 31B26.5%
    Source
    Grok 4.6—

    Not directly comparable

  • HLE w/o tools

    Gemma 4 31B19.5%
    Source
    Grok 4.6—

    Not directly comparable

  • GPQA Diamond (Vals)

    Gemma 4 31B—
    Grok 4.694.7%
    Source

    Not directly comparable

  • MMLU-Pro (Vals)

    Gemma 4 31B—
    Grok 4.689.4%
    Source

    Not directly comparable

25 public results · 0 shared

Watch Gemma 4 31B vs Grok 4.6

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

Read a sample issue

Join 2,000+ readers.

Last updated September 25, 2026