Gemma 4 31B vs GPT-5.6 Luna

Updated September 24, 2026. Rank says GPT-5.6 Luna is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Decision reading

GPT-5.6 Luna has the higher public score estimate, 65.6 versus 44.84, but the 90% score intervals overlap. Treat that as a lead, not a settled winner. 2 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A

Gemma 4 31B

Google

44.84/100

Estimated · Public rank #97

90% interval 24.3–65.3

Model B

GPT-5.6 Luna

OpenAI

65.6/100

Supported · Public rank #25

90% interval 60.5–70.7

Shared results: 2
Gemma 4 31B only: 6
GPT-5.6 Luna only: 27
Like-for-like categories: 3 / 8

Estimated: Gemma 4 31B · Supported: GPT-5.6 LunaHow 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
GPT-5.6 Luna
GPT-5.6 Luna leads on the public coding lane, 64.5 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
GPT-5.6 Luna
GPT-5.6 Luna 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.

Pick the job you would use them for

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 31B64.5GPT-5.6 Luna

Like-for-like · BenchAlign v5.7

GPT-5.6 Luna leads the like-for-like coding row.

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

Open the coding leaderboard Shortlist models for this job

Same basis rules as the category table below

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.

GPQAKnowledge
Normalized gap 8.0
Gemma 4 31B:84.3%
GPT-5.6 Luna:92.3%
Gemma 4 31B source GPT-5.6 Luna source
MMMU-ProMultimodal
Normalized gap 1.5
Gemma 4 31B:76.9%
GPT-5.6 Luna:78.4%
Gemma 4 31B source GPT-5.6 Luna source

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 v5.7 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
GPT-5.6 Luna: 64.5; Supported · #9/135
Basis: BenchAlign v5.7 lane · 2 vs 7 public rows
Reading: GPT-5.6 Luna leads

Multimodal

Like-for-like

Gemma 4 31B: 58.5; #32/50
GPT-5.6 Luna: 67.1; #22/50
Basis: Provisional lane · 1 vs 1 weighted rows
Reading: GPT-5.6 Luna leads

Knowledge

Like-for-like

Gemma 4 31B: 40.0; Supported · #91/158
GPT-5.6 Luna: 64.6; Supported · #22/158
Basis: BenchAlign v5.7 lane · 4 vs 6 public rows
Reading: GPT-5.6 Luna leads

Agentic

Directional only

Gemma 4 31B: 23.9; Estimated · #83/105
GPT-5.6 Luna: 55.3; Supported · #28/105
Basis: BenchAlign v5.7 lane · 1 vs 9 public rows
Reading: Directional only

Reasoning

Not comparable

Gemma 4 31B: 70.0; Unranked · 2 rankable rows
GPT-5.6 Luna: 54.7; #18/19
Basis: Provisional lane · 0 vs 2 weighted rows
Reading: Not comparable

Multilingual

Not comparable

Gemma 4 31B: Not ranked
GPT-5.6 Luna: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Instruction following

Not comparable

Gemma 4 31B: 91.5; #12/124
GPT-5.6 Luna: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Math

Not comparable

Gemma 4 31B: Not ranked
GPT-5.6 Luna: 94.2; Unranked · 3 rankable rows
Basis: Provisional lane · 0 vs 2 weighted rows
Reading: Not comparable

Category scores with the server-provided evidence basis for Gemma 4 31B and GPT-5.6 Luna
Category	Gemma 4 31B	GPT-5.6 Luna	Basis	Reading
Coding	35.7Supported · #72/135	64.5Supported · #9/135	Like-for-likeBenchAlign v5.7 lane · 2 vs 7 public rows	GPT-5.6 Luna leads
Multimodal	58.5#32/50	67.1#22/50	Like-for-likeProvisional lane · 1 vs 1 weighted rows	GPT-5.6 Luna leads
Knowledge	40.0Supported · #91/158	64.6Supported · #22/158	Like-for-likeBenchAlign v5.7 lane · 4 vs 6 public rows	GPT-5.6 Luna leads
Agentic	23.9Estimated · #83/105	55.3Supported · #28/105	Directional onlyBenchAlign v5.7 lane · 1 vs 9 public rows	Directional only
Reasoning	70.0Unranked · 2 rankable rows	54.7#18/19	Not comparableProvisional lane · 0 vs 2 weighted rows	Not comparable
Multilingual	Not ranked	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Instruction following	91.5#12/124	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Math	Not ranked	94.2Unranked · 3 rankable rows	Not comparableProvisional lane · 0 vs 2 weighted rows	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
GPT-5.6 Luna: $0.0008; 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
GPT-5.6 Luna: $0.0136; 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
GPT-5.6 Luna: $0.02; 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.

SpecificationGemma 4 31BGPT-5.6 Luna

Context window

Maximum documented context; output-token limits may be lower.

Gemma 4 31B

256K

Google Gemma 4 model documentation

GPT-5.6 Luna

1.05M

OpenAI model catalog

API model ID

Gemma 4 31B

gemma-4-31b-it

Google Gemma Gemini API guide

GPT-5.6 Luna

gpt-5.6-luna

OpenAI model catalog

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

GPT-5.6 Luna

$0.02 per 1M cached input tokens

OpenAI pricing

Documented inputs

Gemma 4 31B

text, image

Google Gemma 4 model documentation

GPT-5.6 Luna

text, image

OpenAI model catalog

Documented outputs

Gemma 4 31B

text

Google Gemma 4 model documentation

GPT-5.6 Luna

text

OpenAI model catalog

Provider availability

Gemma 4 31B

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

Google Gemma Gemini API guide

GPT-5.6 Luna

Generally Available · OpenAI Responses API

OpenAI model catalog

Reasoning profile

Gemma 4 31B

Reasoning

GPT-5.6 Luna

Reasoning

Weight access

Gemma 4 31B

Open Weight

GPT-5.6 Luna

Proprietary

License

Gemma 4 31B

Open Weight

GPT-5.6 Luna

Proprietary

Release date

Gemma 4 31B

2026-04-02

GPT-5.6 Luna

2026-07-09

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: GPT-5.6 Luna has the higher public score estimate, 65.6 versus 44.84, but the 90% score intervals overlap.
Workload cost: A complete comparable API-rate estimate is not available for both models.
Context tradeoff: GPT-5.6 Luna 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 31B or GPT-5.6 Luna?

GPT-5.6 Luna has the higher public score estimate, 65.6 versus 44.84, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, Gemma 4 31B or GPT-5.6 Luna?

GPT-5.6 Luna leads the public coding lane, 64.5 to 35.7, with Supported evidence for both models and non-overlapping 90% intervals.

Which is better for agentic tasks, Gemma 4 31B or GPT-5.6 Luna?

GPT-5.6 Luna scores higher for agentic tasks on the public lane, 55.3 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 GPT-5.6 Luna?

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 GPT-5.6 Luna?

GPT-5.6 Luna 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 31B

API / mo$0

Self-host / mo$429

Break-even—

GPT-5.6 Luna

API / mo$1,050

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 evidence35 rows

Agentic

Gert Labs
Gemma 4 31B35.26%
Source
GPT-5.6 Luna—
Not directly comparable
Terminal-Bench 3.0
Gemma 4 31B—
GPT-5.6 Luna14.3%
Source
Not directly comparable
Terminal-Bench 2.1
Gemma 4 31B—
GPT-5.6 Luna84.7%
Source
Not directly comparable
BrowseComp
Gemma 4 31B—
GPT-5.6 Luna83.3%
Source
Not directly comparable
OSWorld 2.0
Gemma 4 31B—
GPT-5.6 Luna45.6%
Source
Not directly comparable
CyberGym
Gemma 4 31B—
GPT-5.6 Luna77.9%
Source
Not directly comparable
ExploitGym
Gemma 4 31B—
GPT-5.6 Luna12.4%
Source
Not directly comparable
Toolathlon
Gemma 4 31B—
GPT-5.6 Luna53.4%
Source
Not directly comparable
Terminal-Bench 2.1 (Vals)
Gemma 4 31B—
GPT-5.6 Luna79.0%
Source
Not directly comparable
ApprenticeBench
Gemma 4 31B—
GPT-5.6 Luna7%
Source
Not directly comparable

Coding

SWE-Rebench
Gemma 4 31B41.6%
Source
GPT-5.6 Luna—
Not directly comparable
React Native Evals
Gemma 4 31B75.2%
Source
GPT-5.6 Luna—
Not directly comparable
SWE-bench Pro
Gemma 4 31B—
GPT-5.6 Luna62.7%
Source
Not directly comparable
Terminal-Bench 2.1
Gemma 4 31B—
GPT-5.6 Luna84.7%
Source
Not directly comparable
DeepSWE
Gemma 4 31B—
GPT-5.6 Luna67.2%
Source
Not directly comparable
FrontierCode 1.1 Extended
Gemma 4 31B—
GPT-5.6 Luna55.1%
Source
Not directly comparable
cursorBench32
Gemma 4 31B—
GPT-5.6 Luna61.1%
Source
Not directly comparable
VulcanBench v3
Gemma 4 31B—
GPT-5.6 Luna85.5%
Source
Not directly comparable
SWE-bench (Vals)
Gemma 4 31B—
GPT-5.6 Luna93.0%
Source
Not directly comparable

Reasoning

ARC-AGI-2
Gemma 4 31B—
GPT-5.6 Luna59.5%
Source
Not directly comparable
ARC-AGI-3
Gemma 4 31B—
GPT-5.6 Luna0.2%
Source
Not directly comparable

Multimodal

MMMU-Pro
Gemma 4 31B76.9%
Source
GPT-5.6 Luna78.4%
Source
GPT-5.6 Luna leads this result
MMMU-Pro w/ Python
Gemma 4 31B—
GPT-5.6 Luna79.5%
Source
Not directly comparable

Knowledge

GPQA
Gemma 4 31B84.3%
Source
GPT-5.6 Luna92.3%
Source
GPT-5.6 Luna leads this result
MMLU-Pro
Gemma 4 31B85.2%
Source
GPT-5.6 Luna—
Not directly comparable
HLE
Gemma 4 31B26.5%
Source
GPT-5.6 Luna—
Not directly comparable
HLE w/o tools
Gemma 4 31B19.5%
Source
GPT-5.6 Luna—
Not directly comparable
GPQA-D
Gemma 4 31B—
GPT-5.6 Luna92.3%
Source
Not directly comparable
HealthBench Professional
Gemma 4 31B—
GPT-5.6 Luna55.7%
Source
Not directly comparable
HealthBench Hard
Gemma 4 31B—
GPT-5.6 Luna32.0%
Source
Not directly comparable
GPQA Diamond (Vals)
Gemma 4 31B—
GPT-5.6 Luna91.7%
Source
Not directly comparable
MMLU-Pro (Vals)
Gemma 4 31B—
GPT-5.6 Luna86.0%
Source
Not directly comparable

Math

FrontierMath (legacy)
Gemma 4 31B—
GPT-5.6 Luna78.6%
Source
Not directly comparable
FrontierMath v2 (Tiers 1-3)
Gemma 4 31B—
GPT-5.6 Luna78.600%
Source
Not directly comparable
FrontierMath v2 (Tier 4)
Gemma 4 31B—
GPT-5.6 Luna58.500%
Source
Not directly comparable

35 public results · 2 shared

Related comparisons

Compare API pricing Read the methodology Open the model selector

Watch Gemma 4 31B vs GPT-5.6 Luna

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 24, 2026

Gemma 4 31B vs GPT-5.6 Luna

Decision reading

Which one for your work

Coding work

Long documents

Agentic work

Chat turn cost

Cache-heavy agent loop cost

Repository review cost

Which one for a specific job

What is actually comparable

Shape of the matched evidence

Category results, on a stated basis

Coding

Multimodal

Knowledge

Agentic

Reasoning

Multilingual

Instruction following

Math

What each workload costs

Chat turn

Repository review

Cache-heavy agent loop

Specification differences

Context window

API model ID

Cached-input rate

Documented inputs

Documented outputs

Provider availability

Reasoning profile

Weight access

License

Release date

If you already use one of these models

Questions

Which is better, Gemma 4 31B or GPT-5.6 Luna?

Which is better for coding, Gemma 4 31B or GPT-5.6 Luna?

Which is better for agentic tasks, Gemma 4 31B or GPT-5.6 Luna?

Which costs less, Gemma 4 31B or GPT-5.6 Luna?

Which has the larger context window, Gemma 4 31B or GPT-5.6 Luna?

Self-host vs API cost

Benchmark evidence

Agentic

Coding

Reasoning

Multimodal

Knowledge

Math

Related comparisons

Watch Gemma 4 31B vs GPT-5.6 Luna