Google: Gemma 4 26B A4B
google/gemma-4-26b-a4b-it
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Google's Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts model whose model card reports 25.2B total parameters with only 3.8B active per token, aiming for faster inference than a dense model of similar size. Sources describe text and image input (OpenRouter's catalog also lists video), a 256K-token context window, multilingual support, optional thinking, function calling, and system prompts. Benchmark figures are vendor-reported; independent third-party evidence is thin.[1][2]
Strengths
- Sparse MoE design: the card reports only 3.8B of 25.2B parameters active per token, which Google says yields inference speed close to a 4B model.[1][2]
- Large 256K-token context window with reported text and image input; the OpenRouter catalog additionally lists video input for this variant.[1][2]
- Reported features include an optional thinking mode, native function calling, system-prompt support, and variable-resolution image processing.[1][2]
Limitations
- No audio input: the card reserves audio for the E2B, E4B, and 12B variants.[2]
- The vendor's own multi-needle retrieval test showed moderate accuracy within the context window, so effective long-context use may trail the 256K specification.[2]
- Evidence is mostly first-party plus OpenRouter catalog aggregates; card benchmark figures are not third-party-verified, knowledge cutoff is unstated, and real workloads need validation.[1][2]
Where to use it
- Document and image understanding such as OCR, chart reading, and PDF or UI parsing, where the card reports dedicated vision features.[2]
- Tool-using and agentic applications, given native function calling plus catalog support for tools and structured outputs; run a pilot to confirm fit for autonomous tasks.[1][2]
- Multilingual assistants or long-document workflows that benefit from large context and optional reasoning, provided long-range retrieval quality is validated.[2]
Sources
- [1] OpenRouter catalog · Google: Gemma 4 26B A4B Sources checked: Nov 13, 2026
- [2] Model documentation · Google: Gemma 4 26B A4B Sources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at...
Specifications
- Model ID
- google/gemma-4-26b-a4b-it
- Canonical ID
- google/gemma-4-26b-a4b-it-20260403
- Creator
- Added to OpenRouter
- 03.04.2026
- Context length
- 262,144
- Modality
- text+image+video->text
- Input modalities
- image, text, video
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.090 / out $0.300
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | reasoning: optional · parameters: include_reasoning, reasoning |
| Coding | CC — below the B band | — |
| Agents | CC — below the B band | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · parameters: include_reasoning, reasoning |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | input: video |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Coding index | 39.3 | CC — below the B band | 13.11.2026 |
| Artificial Analysis | Agentic index | 2.9 | CC — below the B band | 13.11.2026 |
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Darkbloom | 131,072 | $0.040 | $0.150 | — | — | 100% |
| DekaLLMbf16 | 262,144 | $0.060 | $0.330 | — | — | 100% |
| PrimeIntellectbf16 | 262,144 | $0.066 | $0.198 | — | — | 100% |
| Io Netbf16 | 262,142 | $0.067 | $0.290 | — | — | 100% |
| AkashMLfp4 | 131,072 | $0.070 | $0.240 | — | — | 100% |
| DeepInfrafp8 | 262,144 | $0.070 | $0.340 | — | — | 100% |
| Makora | 256,000 | $0.090 | $0.280 | — | — | 100% |
| NextBitbf16 | 262,144 | $0.090 | $0.300 | — | — | 100% |
| CoreWeavebf16 | 262,144 | $0.100 | $0.300 | — | — | 100% |
| Cloudflare | 256,000 | $0.100 | $0.300 | — | — | 100% |
| Venicebf16 | 256,000 | $0.130 | $0.400 | — | — | 100% |
| Parasailbf16 | 262,144 | $0.130 | $0.400 | — | — | 100% |
| Novitabf16 | 262,144 | $0.130 | $0.400 | — | — | 100% |
| SiliconFlowfp8 | 262,144 | $0.140 | $0.400 | — | — | 100% |
| 262,144 | $0.150 | $0.600 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.