Baltor Get started

Models

gemma-4-26B-A4B-it

By google. 25.8 billion parameters, a context window of 262,144 tokens and the licence apache-2.0.

Facts

Released
2026-03-11 the day the repository was first published on Hugging Face Hugging Face, read
Licence
apache-2.0 the licence the model card declares Hugging Face, read
Open weights
Yes the weights are published in this Hugging Face repository Hugging Face, read
Parameters
25.8 billion counted from the safetensors weight files Hugging Face, read
Active parameters
Unknown
Knowledge cutoff
Unknown
Context window
  • 262,144 tokens the maximum position embeddings in the model configuration model configuration, read
Longest output
Unknown
Tool calling
Unknown
Structured output
Unknown
Reasoning controls
Unknown
Inputs and outputs
text, image in, text out Hugging Face, read
Good for
  • Vision: Hugging Face files it under image-text-to-text Hugging Face, read

Prices

Each provider's own price in US dollars per million tokens, as an openly licensed source lists it
ProviderInputOutputContextAs ofSource
Kilo Gateway0.042 USD0.220 USD262,144 older than 30 days; check the providermodels.dev, read
OrcaRouter0.060 USD0.330 USD262,144 older than 30 days; check the providermodels.dev, read
Deep Infra0.070 USD0.340 USD262,144 older than 30 days; check the providermodels.dev, read
NanoGPT0.090 USD0.300 USD262,144 older than 30 days; check the providermodels.dev, read
OpenRouter0.090 USD0.300 USD262,144 older than 30 days; check the providermodels.dev, read
CoreWeave0.100 USD0.300 USD262,144 older than 30 days; check the providermodels.dev, read
SiliconFlow0.120 USD0.400 USD262,144 older than 30 days; check the providermodels.dev, read
Hugging Face0.130 USD0.400 USD262,144 older than 30 days; check the providermodels.dev, read
Merge Gateway0.130 USD0.400 USD262,144 older than 30 days; check the providermodels.dev, read
Novita AI0.130 USD0.400 USD262,144 older than 30 days; check the providermodels.dev, read
evroc0.144 USD0.575 USD262,144 older than 30 days; check the providermodels.dev, read
Vercel AI Gateway0.150 USD0.600 USD262,144 older than 30 days; check the providermodels.dev, read
AI/ML API0.206 USD0.825 USD262,144 older than 30 days; check the providermodels.dev, read
Scaleway0.250 USD0.500 USD256,000LiteLLM, read

Run it on your own hardware

Weights are the sizes of the files a source lists, or an estimate from the parameter count where none does. Memory is an estimate: weights plus KV cache plus 512 MiB and 5 percent of the weights for runtime buffers. Check it against your hardware.

QuantizationWeightsMemory at 8,192 tokensMemory at 32,768 tokens
UD-IQ2_XXS9.2 GiB12.1 GiB17.7 GiB
UD-IQ2_M9.3 GiB12.2 GiB17.8 GiB
UD-Q2_K_XL9.8 GiB12.7 GiB18.3 GiB
UD-IQ3_S10.5 GiB13.4 GiB19.0 GiB
UD-IQ3_XXS10.6 GiB13.5 GiB19.2 GiB
UD-Q3_K_M11.9 GiB14.8 GiB20.4 GiB
UD-Q3_K_XL12.0 GiB15.0 GiB20.6 GiB
UD-IQ4_XS12.7 GiB15.7 GiB21.3 GiB
UD-IQ4_NL12.7 GiB15.7 GiB21.3 GiB
UD-Q4_K_S15.4 GiB18.5 GiB24.1 GiB
MXFP4_MOE15.4 GiB18.6 GiB24.2 GiB
UD-Q4_K_M15.8 GiB18.9 GiB24.6 GiB
UD-Q4_K_XL15.8 GiB19.0 GiB24.6 GiB
UD-Q5_K_S17.6 GiB20.8 GiB26.4 GiB
UD-Q5_K_M19.7 GiB23.1 GiB28.7 GiB
UD-Q5_K_XL19.8 GiB23.1 GiB28.7 GiB
UD-Q6_K21.6 GiB25.0 GiB30.7 GiB
UD-Q6_K_XL21.7 GiB25.2 GiB30.8 GiB
Q8_025.4 GiB29.1 GiB34.7 GiB
UD-Q8_K_XL25.7 GiB29.4 GiB35.0 GiB

Start it with a local runtime

  • Ollama
    ollama run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ2_XXS
  • LM Studio
    lms get unsloth/gemma-4-26B-A4B-it-GGUF
    lms server start
  • llama.cpp server
    llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ2_XXS --jinja
  • vLLM
    vllm serve google/gemma-4-26B-A4B-it
  • SGLang
    sglang serve --model-path google/gemma-4-26B-A4B-it --port 30000

Use it from your harness

Set up for AI/ML API with the model google/gemma-4-26b-a4b-it. Each endpoint page has the same setup for its own address.

OpenCode

Put this in opencode.json in your project folder:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "aimlapi": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "AI/ML API",
      "options": {
        "baseURL": "https://api.aimlapi.com/v1",
        "apiKey": "{env:AIMLAPI_API_KEY}"
      },
      "models": {
        "google/gemma-4-26b-a4b-it": {
          "name": "gemma-4-26B-A4B-it"
        }
      }
    }
  }
}
  • OpenCode reads any OpenAI-compatible address through the @ai-sdk/openai-compatible package, and an address that speaks the Responses API through @ai-sdk/openai.

From OpenCode documentation, read .

Pi

Put this in ~/.pi/agent/models.json:

{
  "providers": {
    "aimlapi": {
      "baseUrl": "https://api.aimlapi.com/v1",
      "api": "openai-completions",
      "apiKey": "$AIMLAPI_API_KEY",
      "models": [
        {
          "id": "google/gemma-4-26b-a4b-it"
        }
      ]
    }
  }
}
  • The apiKey field can name an environment variable as $NAME.

From Pi documentation, read .

Codex

Codex speaks only the Responses API, and AI/ML API documents no Responses address. A gateway that offers one can sit in between.

  • Codex speaks the Responses API only: responses is the one supported wire API of a custom provider. Ollama and LM Studio are built in and start with --oss.

From Codex documentation, read .

Claude Code

Claude Code sends Anthropic Messages requests, and AI/ML API documents no such address. A gateway that translates to that API can sit in between.

  • Claude Code sends Anthropic Messages requests to ANTHROPIC_BASE_URL. Anthropic says it does not support routing Claude Code to models other than Claude through any gateway, so some features may not work with another model.

From Claude Code documentation, read .

Other ways to reach it

Published results

These are results other people published. Baltor did not run them and does not rank models by them.

Two more publishers are linked, never copied, because their terms do not allow it:

  • Artificial Analysis publishes its own intelligence, speed and price measurements.
  • OpenRouter lists current prices and uptime for each provider it routes to.