Gemma 4 26B-A4B vs Gemma 4 31B

Why this pair: Largest Gemma 4 instruction-tuned variants

Benchmark Performance

Available Benchmarks

BenchmarkGemma 4 26B-A4BGemma 4 31B
LMArena Text Arenatext-2026-09-01-011508720696 · arena_rating · statistical tie1,434.86100% of row best · rating · gemma-4-26b-a4b; 95% CI [1427.26766156, 1442.45594559]; votes 5814; rank 781,441.82100% of row best · rating · gemma-4-31b; 95% CI [1434.27306811, 1449.35895578]; votes 5900; rank 60
LMArena Vision Arenavision-2026-08-27-011508720696 · arena_rating · leader1,259.4599% of row best · rating · gemma-4-26b-a4b; 95% CI [1252.74042036, 1266.15761735]; votes 22710; rank 431,275.94100% of row best · rating · gemma-4-31b; 95% CI [1269.53471972, 1282.35229256]; votes 34348; rank 31
Overall ResultCounted from the protocol-matched rows above · 1 tie0 benchmark wins1 benchmark winOverall lead

Third-party benchmark Only like-for-like primary-publisher results are shown; raw scores, relative scores, configuration, and token spend remain visible.

FieldAt a Glance
Google Deepmind · activeGemma 4 26B-A4BVerified Sep 3, 2026
Google Deepmind · activeGemma 4 31BVerified Sep 3, 2026
Not a Valid ComparisonThese records do not share a sourced modality and workload.

Technical Differences

Side-by-Side Facts

CuratedInteractive only
FieldGemma 4 26B-A4BGemma 4 31B
DeveloperGoogle DeepmindGoogle Deepmind
FamilyGemma 4Gemma 4
ModelGemma 4 26B-A4BGemma 4 31B
VersionGemma 4 26B-A4BGemma 4 31B
Lifecycleactiveactive
Released2026-03-112026-03-11
Knowledge cutoffUnknownUnknown
Input modalitiesUnknownUnknown
Output modalitiesUnknownUnknown
Context window262,144262,144
Total parameters26,000,000,00031,000,000,000
Active parameters4,000,000,000Unknown
Licenseapache-2.0apache-2.0
Open weightsYesYes
API availableYesYes
Self-hostableYesYes
Provider accessDeepinfra (Standard), Google Gemini (Standard), Openrouter (Standard), Together Ai (Standard)Cerebras (Standard), Deepinfra (Standard), Google Gemini (Standard), Openrouter (Standard), Together Ai (Standard)
Capabilitieschat, generation, reasoning, structured_outputs, toolschat, generation, reasoning, structured_outputs, tools

14 comparable fields · 4 material differences · Interactive comparison only; indexing gate not met

Gemma 4 26B-A4B Capabilities

chatgenerationreasoningstructured outputstools
Input price$0.070
Output price$0.34
Serving providers4
Canonical IDgoogle/gemma-4-26B-A4B-it

Gemma 4 31B Capabilities

chatgenerationreasoningstructured outputstools
Input price$0.090
Output price$0.34
Serving providers5
Canonical IDgoogle/gemma-4-31B-it

Primary Evidence

Sources and Freshness

Questions

Gemma 4 26B-A4B vs Gemma 4 31B FAQs

Is Gemma 4 26B-A4B or Gemma 4 31B better for coding?+

This comparison does not currently contain a protocol-matched coding benchmark for both Gemma 4 26B-A4B and Gemma 4 31B, so Model Markets cannot name a coding leader from pricing, context size, or capability labels alone.

Which is cheaper, Gemma 4 26B-A4B or Gemma 4 31B?+

Gemma 4 26B-A4B is $0.070 and Gemma 4 31B is $0.090 per million tokens, so Gemma 4 26B-A4B is cheaper on this metric. Gemma 4 26B-A4B is $0.34 and Gemma 4 31B is $0.34 per million tokens, so they are tied on this metric.

Which has a larger context window, Gemma 4 26B-A4B or Gemma 4 31B?+

Neither model has a larger sourced context window in this comparison. Gemma 4 26B-A4B is 262,144 and Gemma 4 31B is 262,144.

Which performs better in benchmarks, Gemma 4 26B-A4B or Gemma 4 31B?+

There is no overall benchmark winner: An overall winner requires at least two decisive benchmarks from at least two original publishers.

Can Gemma 4 26B-A4B or Gemma 4 31B be self-hosted?+

Both models have the same recorded self-hosting status: supported. Gemma 4 26B-A4B is open weight; Gemma 4 31B is open weight.

Can Gemma 4 26B-A4B and Gemma 4 31B understand images?+

Gemma 4 26B-A4B is not documented with image input; Gemma 4 31B is not documented with image input. This reflects supported input modalities, not vision quality.

Which can generate longer answers, Gemma 4 26B-A4B or Gemma 4 31B?+

Neither has a larger sourced maximum output. Gemma 4 26B-A4B is — and Gemma 4 31B is —.

Do Gemma 4 26B-A4B and Gemma 4 31B support reasoning and tool use?+

Gemma 4 26B-A4B: reasoning and tool calling. Gemma 4 31B: reasoning and tool calling. Feature support does not establish relative quality.

Which is available from more inference providers, Gemma 4 26B-A4B or Gemma 4 31B?+

Gemma 4 26B-A4B has 4 sourced provider routes; Gemma 4 31B has 5, so Gemma 4 31B has broader tracked availability.

Which offers better value, Gemma 4 26B-A4B or Gemma 4 31B?+

There is no universal value winner. Compare the input and output prices above with the matched benchmark result for your workload: cheaper tokens can be offset by different quality, token usage, latency, or provider availability.

Send Feedback