Gemma
Google's open models — Gemini's research, small enough to self-host.
Gemma is Google DeepMind's open-weight family, built from the same research as Gemini but published so you can download and run it. Sizes run from a few billion parameters up to about thirty, they read text and images, and the current line has a 256K context with a switchable thinking mode. The obvious starting point if you want a capable model on your own machine or on cheap hardware.
At a glance — Gemma 4 31B
- Cost
- $0.09 in · $0.34 out per 1M tokensBudgetFree version too
- Memory
- 262K tokens — about 400 pages
- Reads
- Text · Images · Video
- Can do
- Use tools (agents)
- Structured output
- Thinking mode
- Good at
- Coding: better than 42% · Agentic work: better than 31%
- Reliability
- 99.94% uptime, last 24h
- Released
- Apr 2, 2026
Advanced details
- API model ID
google/gemma-4-31b-it- Free version
google/gemma-4-31b-it:free- Max output
- 16K tokens
- Cached input
- $0.05 per 1M
- Tokenizer
- Gemma
- Coding index
- 43.4
- Agentic work index
- 4.2
Providers
The same model, hosted by different companies. OpenRouter picks one per request and falls back to the next if it fails.
| Provider | Input /1M | Output /1M | Cached /1M | Max output | Uptime 24h |
|---|---|---|---|---|---|
| RekaDegraded | $0.09 | $0.34 | $0.05 | 131K | 94.50% |
| DeepInfra · turbo | $0.09 | $0.34 | $0.05 | 16K | 98.87% |
| CoreWeave · fp4 | $0.10 | $0.34 | $0.10 | 236K | 98.26% |
| Venice · fp4 | $0.12 | $0.36 | $0.09 | 8K | 99.04% |
| Chutes · fp4 | $0.12 | $0.37 | $0.012 | 66K | 94.57% |
| DeepInfra · fp8 | $0.13 | $0.38 | — | 16K | 98.91% |
| Crusoe · bf16Degraded | $0.14 | $0.40 | $0.14 | 262K | 99.29% |
| Friendli | $0.14 | $0.40 | — | 8K | 99.93% |
| Novita · bf16Degraded | $0.14 | $0.40 | — | 131K | 98.74% |
| Parasail · fp8 | $0.15 | $0.40 | $0.06 | 236K | 98.54% |
| DeepInfra · ultra | $0.27 | $0.76 | — | 8K | 90.52% |
| Io Net | $0.38 | $1.15 | $0.19 | 16K | 98.53% |
| SambaNova | $0.38 | $1.15 | — | 118K | 95.58% |
| ModelRun · fp4 | $0.75 | $1.00 | $0.75 | 236K | 99.94% |
| SiliconFlow · fp8 | $0.75 | $1.00 | $0.25 | 236K | 92.61% |
Supported parameters
- frequency_penalty
- include_reasoning
- logit_bias
- logprobs
- max_tokens
- min_p
- presence_penalty
- reasoning
- repetition_penalty
- response_format
- seed
- stop
- structured_outputs
- temperature
- tool_choice
- tools
- top_k
- top_logprobs
- top_p
Live from OpenRouter, refreshed hourly. Scores from Artificial Analysis.
Key info
- Pricing
- Open source
- Category
- Models
- Best for
- Intermediate
- Made by
- Google DeepMind
- Weights
- Open
- Good for
- Running a capable model on your own machine