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 26B A4B
- Cost
- $0.068 in · $0.225 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 38%
- Reliability
- 99.98% uptime, last 24h
- Released
- Apr 3, 2026
Advanced details
- API model ID
google/gemma-4-26b-a4b-it- Free version
google/gemma-4-26b-a4b-it:free- Max output
- 236K tokens
- Cached input
- $0.038 per 1M
- Tokenizer
- Gemma
- Coding index
- 39.3
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 |
|---|---|---|---|---|---|
| Darkbloom | $0.042 | $0.22 | — | 33K | 99.98% |
| DekaLLM · bf16 | $0.06 | $0.33 | — | 236K | 99.91% |
| NextBit · bf16 | $0.068 | $0.225 | $0.038 | 236K | 99.87% |
| DeepInfra · fp8 | $0.07 | $0.34 | — | 16K | 99.79% |
| Makora | $0.08 | $0.32 | $0.032 | 128K | 99.61% |
| Reka | $0.09 | $0.30 | $0.05 | 131K | 99.32% |
| CoreWeave · bf16 | $0.10 | $0.30 | $0.05 | 236K | 99.83% |
| Cloudflare | $0.10 | $0.30 | — | 230K | 99.91% |
| Venice · bf16 | $0.13 | $0.40 | $0.05 | 8K | 99.27% |
| Parasail · bf16 | $0.13 | $0.40 | $0.05 | 236K | 99.25% |
| Novita · bf16 | $0.13 | $0.40 | — | 131K | 99.89% |
| SiliconFlow · fp8 | $0.14 | $0.40 | $0.05 | 236K | 97.54% |
| Io Net · bf16 | $0.15 | $0.50 | $0.15 | 236K | 99.19% |
| Google · globalDegraded | $0.15 | $0.60 | — | 236K | 96.22% |
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