Gemma4 26B A4B Uncensored (Balanced)

obsidian/gemma-4-26B-A4B
New
VisionToolsJSONReasoning
by obsidian · 2026-07-02

A lossless uncensored version of Gemma 4 26B A4B designed to preserve the original model's capabilities while minimizing refusal behavior. Optimized for developers, AI researchers, and advanced applications that require high-quality responses with minimal restrictions. The Balanced variant is recommended for most workloads, delivering complete answers while maintaining stable reasoning and natural conversational behavior. In some sensitive scenarios, the model may briefly frame its reasoning before providing the full response, but it is designed to avoid withholding content. Compared to more aggressive uncensored variants, Balanced offers more consistent sampling, stronger long-context stability, and reduced topic drift across extended conversations. Well suited for creative writing, roleplay, multilingual assistants, long-context reasoning, and general-purpose AI applications where quality, coherence, and reliability are the primary priorities. Access to this model is gated and intended for security researchers, red teams, AI safety researchers, and other qualified professionals conducting legitimate research, evaluation, and testing.

ctx262.1K tokens
Inputtext + image
Outputtext
Best foruncensored, reasoning, vision
p50 TTFT3.50 s
INPUT$0.25/ 1M tokens
OUTPUT$2.90/ 1M tokens
p50 TTFT3.50 s7d
p95 TTFT4.13 s7d
TRAFFIC60.8Ktokens / 7d

Gemma 4 26B A4B Uncensored is a mixture-of-experts (MoE) language model from Google’s Gemma 4 series, hosted by provider Obsidian and accessible through OrcaRouter. It has 26 billion total parameters…

What is Gemma 4 26B A4B Uncensored?

Who should use this model?

What input modalities does it support?

How does the uncensored version differ from the base Gemma 4?

Pricing

Input / 1M tokens$0.250
Output / 1M tokens$2.90
CurrencyUSD

Cost calculator

Tokens / month10MM
Input share70%%
Estimated / month $10.45

Estimate based on list price

Token & cost estimator

Input tokens: 20Cost per request: $0.001455

Estimate only — actual token counts depend on the provider's tokenizer.

Performance

p50 TTFT
3.50 s
Output speed
225 tok/s
p95 TTFT
4.13 s
Error rate
0%

Public benchmarks

39.3
AA Coding
Better than 44% of models compared
#66 of 120
25.7
AA Intelligence
Better than 31% of models compared
#83 of 122
GPQA Diamond
79.2
Humanity's Last Exam
18.3
IFBench
72.4
Long-Context Recall
55.7
SciCode
40.0
tau_banking
11.8
TerminalBench Hard
13.6
terminalbench_v2_1
39.0
τ²-Bench
43.6
Source: artificialanalysis.ai

How it compares

Gemma4 26B A4B Uncensored (Balanced)Qwen3.6 35B A3B Uncensored (Aggressive)
Input $/M$0.25$0.31
Output $/M$2.90$4.21
Context262K262K
Quality4/104/10
Compare side-by-sideCompare side-by-side

FAQ

What is the cost per token for this model?
The cost is $0.25 per 1 million input tokens and $2.90 per 1 million output tokens. There is no markup from OrcaRouter; this is the provider rate.
What is the context window size?
The context window is 262,144 tokens. This includes both input and output tokens. The window is shared between text and image tokens.
What are the strengths of this model?
Its strengths include a large context window, MoE efficiency (4B active parameters), multimodal input, and uncensored output. It is good for long-document tasks and unrestricted content generation.
How does it compare to other Gemma models?
Compared to dense Gemma models of similar total size (e.g., Gemma 2 27B), it is more computationally efficient. The uncensored version removes safety filters.
Does the model support image inputs?
Yes, it accepts both text and image inputs. Images can be passed as URLs or base64 encodings within the API request.
How do I call this model via OrcaRouter?
Use base URL https://api.orcarouter.ai/v1 and model ID "obsidian/gemma-4-26B-A4B". Include your API key in the Authorization header.
What data handling policies apply?
Data handling is governed by OrcaRouter's privacy policy and the provider Obsidian. As this model is uncensored, users are responsible for ensuring compliance with applicable laws and terms of service.
Is there any caching or discount for repeated prompts?
OrcaRouter does not offer built-in caching for this model. No discounts for repeated prompts are listed. You may implement client-side caching.
What parameters can I use with the API?
Standard OpenAI-compatible parameters: temperature, top_p, max_tokens, stop, presence_penalty, frequency_penalty, n, and stream. Multimodal content uses the image_url format.
Can I use this model for commercial applications?
Yes, you can use it commercially, but you must ensure your use case complies with all applicable laws and OrcaRouter's terms. The uncensored nature may require additional safety measures.

Embed this badge

Gemma4 26B A4B Uncensored (Balanced)$0.25/M in3500ms p50via OrcaRouter
HTML <a href="https://www.orcarouter.ai/models/obsidian/gemma-4-26B-A4B" target="_blank"> <img src="https://www.orcarouter.ai/embed/obsidian/gemma-4-26B-A4B.svg" alt="Gemma4 26B A4B Uncensored (Balanced) on OrcaRouter" /> </a>
Markdown [![Gemma4 26B A4B Uncensored (Balanced)](https://www.orcarouter.ai/embed/obsidian/gemma-4-26B-A4B.svg)](https://www.orcarouter.ai/models/obsidian/gemma-4-26B-A4B)

Model card as data

GET /api/public/models/obsidian/gemma-4-26B-A4BOpen
Machine-readable:/llms.txt/llms-full.txt