Models

The barsare off.

Open-weight models with the refusal behaviour removed from the weights. The prices below come straight from the table that bills you.

2 modelsrefusals ablatedno prompt logs
Try in chat API docs

Pick the thinker.Or the sprinter.

Both speak the OpenAI API. Both stream, call tools and return JSON. They differ in what wakes up for each token, and in what that costs.

Qwen · 27B dense
venym/qwen-27b

Qwen 27BUnfiltered

The thinker. Dense 27B, refusals removed.

One token wakes every weight
Active per token 27B of 27B (100%)
Context
131.1K
Max output
32.8K
Quantization
FP8
Architecture
Transformer, GQA
Input / 1M
$0.30
Output / 1M
$1.20
chattoolsjsonreasoningweb
Spec sheet Try in chat
GLM · 320B MoE, 18B active
venym/glm-flash

GLM 5.3 FlashUnfiltered

The sprinter. Fast MoE, no leash.

One token wakes only this slice
Active per token 18B of 320B (5.6%)
Context
131.1K
Max output
16.4K
Quantization
FP8
Architecture
Mixture of experts
Input / 1M
$0.10
Output / 1M
$0.40
chattoolsjsonweb
Spec sheet Try in chat

Side by side.Tell it what you need.

Pick what matters. The column that fits lights up, with the rows that decided it.

I need
Model comparison
Qwen
qwen-27b
best fit
GLM
glm-flash
best fit
Parameters
27B dense
320B MoE, 18B active
Active per token
27B100% of all
18B5.6% of all
Architecture
Transformer, GQA
Mixture of experts
Quantization
FP8
FP8
Context window
131.1K
131.1K
Max output
32.8K
16.4K
Input / 1M
$0.30
$0.10
Output / 1M
$1.20
$0.40
Chat
yes
yes
Tool calls
yes
yes
JSON mode
yes
yes
Reasoning
yes
—
Web (:online)
yes
yes

Reasoning is a listed capability of the model.

venym/qwen-27b:online

One suffix on the model id, or "venym": { "web_search": true } in the body. A search runs first and the answer cites its sources.

POST /v1/chat/completions
online
Web search on
Request
POST /v1/chat/completions

{
  "model": "venym/qwen-27b:online",
  "messages": [
    {
      "role": "user",
      "content": "What is the latest stable release of Example Framework?"
    }
  ]
}
Response · illustrationgrounded
Search ran first · 3 results$0.0020
[1]docs.example.com/releases[2]example.com/blog/release-4-2[3]docs.example.com/migrate
Search call
$0.0020
1 × $0.0020
Prompt tokens
$0.0004
1,374 × $0.30 / 1M
Completion
$0.0001
74 × $1.20 / 1M
This call
$0.0025
search + extra prompt tokens
Question, sources and token counts are illustrativeDocs
Ready when you are

$0.25 on the house.No email to give.

Get an account See pricing