Gemma 4 Abliterated API: 26B at $0.25/$0.75 with Vision

Gemma 4 Abliterated API pricing on RouterPlex is $0.25 per 1M input tokens and $0.75 per 1M output tokens, and the model id is gemma-4-26b-abliterated. Cached input bills at the input rate. Checked 3 November 2026. Open Gemma 4 26B Abliterated for the live rate.

This is Google's Gemma 4 26B, fine-tuned so it stops refusing. It is the only abliterated route on RouterPlex that combines image input, tool calls, and fast replies with no reasoning step. RouterPlex sets this rate, and it is not a Google list price.

RouterPlex publishes this page and sells the prepaid wallet described below.

Create a free RouterPlex account and add a $5 top-up. Cap the key, then send one short request before you move a real job onto this id.

A fine-tune, not a weight edit

There are two common ways to take refusals out of an open model. Abliteration edits the weights directly and needs no more training. That is how the other four routes in this set were made, and what abliteration is explains it. A fine-tune trains the model further on examples where it answers instead of refusing.

gemma-4-26b-abliterated is a fine-tune of Gemma 4 26B-A4B, a mixture-of-experts model with 26B total parameters and about 4B active per token. The small active size is why it answers quickly. It is served quantized to 4 bits, so expect it to be weaker than the full-precision Gemma 4 on hard reasoning. RouterPlex lists it with the abliterated models because it does the same job: refusals are removed.

What it is for

Stock models often refuse legitimate work on sensitive material. This route answers it, and it can read images while it does. Incident responders paste a screenshot of a phishing page or a suspicious document and ask what it does. Trust and safety teams label harmful images and text for their own filters. Red teams test their own LLM apps with attack prompts, including prompts hidden in images. Writers draft fiction for adult readers that deals with violence or crime.

You are still bound by the terms. Authorized work is your responsibility, and the model can be wrong, so check what it gives you.

Gemma 4 Abliterated API pricing

USD per 1M tokens, from the RouterPlex catalog on 3 November 2026.

A cold call of 40,000 input tokens and 2,000 output tokens costs $0.0115 here. The same mix is $0.0060 on glm-4.7-flash-abliterated and $0.012 on glm-5.3-flash-abliterated. Gemma bills no reasoning tokens. The other four routes reason before they answer, and that hidden text bills as output.

Cached input is not discounted on this route. Our upstream counts cached tokens here (a repeated 8,033-token prompt came back with 8,032 cached) but does not bill them at a lower rate, so neither do we. For prompts you resend many times, glm-5.3-flash-abliterated at $0.06 cached is cheaper.

A $5 balance covers 20 million input tokens at $0.25, or about 6.7 million output tokens at $0.75.

Our refusal check

On 3 November 2026 we sent five prompts that stock models often decline: an explicit scene between two adult characters, a graphic horror scene, SQL injection payloads for a login form, an aggressive essay against religion, and a request for a vicious personal insult. Gemma 4 Abliterated answered all five. As a control, stock GLM-5.3 Flash refused two of the five and GPT-6 Luna refused three.

Five prompts is a small check, not a benchmark. It shows the route answers the kind of request a stock model declines. It does not measure answer quality. Run your own evaluation set before you depend on it.

Speed

With no reasoning step, the answer starts right away. In the refusal check, answers of up to 2,500 tokens took about 10 seconds on average. GLM-4.7 Flash Heretic took about 26 seconds on the same prompts.

Specs

Context is 256,000 tokens. Max output is 8,192, so long reports need to be split into parts. Image input works through image_url parts. We sent a solid red test image and the model answered "Red". Tool calls work, including streamed tool calls. The public page slug is /models/gemma-4-26b-abliterated.

Setup

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemma-4-26b-abliterated",
    "messages": [
      {"role": "user", "content": [
        {"type": "text", "text": "Describe what this screenshot asks the user to do."},
        {"type": "image_url", "image_url": {"url": "https://example.com/screenshot.png"}}
      ]}
    ]
  }'

Replace the image URL with your own, or send a base64 data:image/png;base64,... URL. Claude Code and other Anthropic clients use the gateway root.

export ANTHROPIC_BASE_URL=https://api.routerplex.com
export ANTHROPIC_AUTH_TOKEN=$ROUTERPLEX_API_KEY
export ANTHROPIC_MODEL=gemma-4-26b-abliterated

The OpenAI-compatible API guide covers the SDK base URL https://api.routerplex.com/v1. Keep the key in the environment. Do not commit it.

Errors

The $1 trial key is not free access to this id, and there is no free credit on abliterated routes. HTTP 402 with type paid_model_required means the key is still a trial key. Add a paid top-up of at least $5, and the top-up fee is $0.

A hard spend cap stops the key even when the account balance remains. Raise the cap or split the job. The account itself does not go negative.

An answer that stops at 8,192 tokens hit the output limit. Ask for the work in parts, or move it to GLM-5.3 Flash Abliterated, which allows 131,072. Model not found means the id string is wrong. Send gemma-4-26b-abliterated, not a Hugging Face repo name.

Which abliterated route to pick

• Fast replies with image input: gemma-4-26b-abliterated, this page.
• Cheapest, text only: GLM-4.7 Flash Heretic, $0.12 and $0.60.
• 1M context with image input: GLM-5.3 Flash Abliterated, $0.25 and $1.00.
• DeepSeek family, 131K output: DeepSeek V4.1 Flash Abliterated, $0.50 and $1.50.
• Strongest reasoning: GLM-5.3 Abliterated, $2 and $4.

All five are listed together in the abliterated section of the model catalog.

Terms

The model will answer some requests that stock Gemma 4 declines, and you are responsible for that use under the terms. Illegal content, malware, weapons development, and child sexual abuse material are barred.

Next step

Create a free RouterPlex account, add a $5 top-up, and open Gemma 4 26B Abliterated before you set a budget.