Gemma 4 31B
The dense Apache-2.0 model local-first teams call 'the new baseline'. Excellent at rule-following, OCR and automation; needs ~48GB for full 256K context and is not a frontier reasoner. Free to run, no API price to speak of.
Key facts
- Score
- 58 / 100, rank #22Evals47Pricen/aPeople72Fit85
- Price
- No verified list price
- Context window
- 256K
- Open weights
- Yes
- Released
- Apr 2, 2026
- Best for
- On-device and workstation inference, OCR and document vision, Deterministic pipeline/automation steps
Google has not published a list price for Gemma 4 31B that we could verify from a primary source.
Independent evals
| Benchmark | Result | Who ran it | Read |
|---|---|---|---|
| Artificial Analysis Intelligence Index | 15 | Gemma 4 31B - Artificial Analysis | Oct 10, 2026 |
| Arena Text score | 1452 (rank 75) | Text Arena Leaderboard | Oct 10, 2026 |
- Artificial Analysis output speed (tokens/s): 34.8 (Gemma 4 31B - Artificial Analysis)
What practitioners say
HN's local-LLM crowd treats it as the reference point: 'the new baseline for local models' (soganess, 70GB peak on an M5 Max at 256K context); 'particularly good at pipeline/automation tasks' and better than Qwen even at 100B+ for rule-following; 'very good at OCR'. One user had it catch a bug Opus 4.7 missed. Counterpoints: Qwen 3.6 'handles context a little better', peers 'predominantly run Qwen', and ~11-16 tok/s on Macs feels slow.
Caveats
- Dense 31B: expect ~12-16 tok/s on Apple Silicon at Q4-Q6, and 70GB RAM at full context.
- Several HN users say Qwen 3.6/3.8 handles long context and agentic tasks a little better; Gemma wins on instruction discipline.
- AA 15 on the current index; this is a local model, not a hosted-API competitor.
Questions people ask
How much does Gemma 4 31B cost?
Google has not published a list price for Gemma 4 31B that we could verify from a primary source.
How good is Gemma 4 31B?
Gemma 4 31B ranks #22 of 26 on the Models at Work leaderboard with a score of 58 out of 100 (October 2026 edition). The dense Apache-2.0 model local-first teams call 'the new baseline'. Excellent at rule-following, OCR and automation; needs ~48GB for full 256K context and is not a frontier reasoner. Free to run, no API price to speak of.
What is Gemma 4 31B best for?
Gemma 4 31B is best for on-device and workstation inference, ocr and document vision, deterministic pipeline/automation steps.
What are Gemma 4 31B's benchmark scores?
Artificial Analysis Intelligence Index: 15 (Gemma 4 31B - Artificial Analysis); Arena Text score: 1452 (rank 75) (Text Arena Leaderboard).
What is Gemma 4 31B's context window?
Gemma 4 31B has a 256K token context window according to Google.
Is Gemma 4 31B open weights?
Yes. Gemma 4 31B is released with open weights, so it can be self-hosted.
Compare Gemma 4 31B
Methodology: Score is 0 to 100 and computed, not typed: 55% independent evals (Artificial Analysis, Arena, Scale SEAL, Epoch, each scaled against its natural floor and the best score in this table, then averaged), 15% blended price on a fixed log scale ($0.05 per million is 100, $60 is 0), 15% practitioner sentiment from Signals and community threads, 15% operational fit (context window, open weights). A missing component drops out and the rest are reweighted. A model with no independent eval yet is provisional and ranks below every measured one. Within three points is a tie. Vendor-published figures are labelled and never counted.