Skip to content
Catalogue · live

Eight models, one endpoint.

5 Qwen releases, two gpt-oss sizes and Llama 3.3 — from a 9B workhorse to a 397B mixture of experts. Same authentication, same streaming contract, same quota.

Pricing
2,500 IQD/ month per model
or 20,000 IQD / year — save 10,000 IQD
See what is included

All models

AvailableNew
01

Qwen3.8

Vision-capable 27B with strong reasoning and tool use.

Plan
2,500 IQD/ month
Use it
idQwen3.8
AvailableNew
02

Qwen3.5

Flagship mixture-of-experts for the hardest reasoning.

Plan
2,500 IQD/ month
Use it
idQwen3.5
AvailableNew
03

Qwen3.6

Previous-generation 27B, kept for evaluation and A/B.

Plan
2,500 IQD/ month
Use it
idQwen3.6
AvailableNew
04

Qwen3.5-Lite

Fast, cheap 9B for chat, extraction and drafting.

Plan
2,500 IQD/ month
Use it
idQwen3.5-Lite
Available
05

gpt-oss-pro

Open-weight 120B mixture-of-experts with strong reasoning per euro.

Plan
2,500 IQD/ month
Use it
idgpt-oss-pro
Available
06

gpt-oss-lite

Small gpt-oss for high-volume chat and classification.

Plan
2,500 IQD/ month
Use it
idgpt-oss-lite
Available
07

Qwen2.5-VL

Vision-language model for documents, screenshots and charts.

Plan
2,500 IQD/ month
Use it
idQwen2.5-VL
Available
08

Llama-3.3

Dependable general-purpose 70B with tool use.

Plan
2,500 IQD/ month
Use it
idLlama-3.3
Side by side

Pick with numbers, not vibes.

Everything is measured in the same units, so you can decide per feature instead of per model.

Comparison of every model in the Kalla catalogue
SpecQwen3.8
Qwen
Qwen3.5
Qwen
Qwen3.6
Qwen
Qwen3.5-Lite
Qwen
gpt-oss-pro
gpt-oss
gpt-oss-lite
gpt-oss
Qwen2.5-VL
Qwen
Llama-3.3
Llama
StatusAvailableNewAvailableNewAvailableNewAvailableNewAvailableAvailableAvailableAvailable
Best forVision-capable 27B with strong reasoning and tool use.Flagship mixture-of-experts for the hardest reasoning.Previous-generation 27B, kept for evaluation and A/B.Fast, cheap 9B for chat, extraction and drafting.Open-weight 120B mixture-of-experts with strong reasoning per euro.Small gpt-oss for high-volume chat and classification.Vision-language model for documents, screenshots and charts.Dependable general-purpose 70B with tool use.
Parameters27B397B27B9.7B117B21B72B70B
Quantizationfp8fp8fp8bf16fp4fp4fp8—
Context window262K tokens262K tokens262K tokens262K tokens131K tokens131K tokens33K tokens131K tokens
CapabilitiesFunction Calling · Reasoning · Streaming · Structured Output · VisionFunction Calling · Reasoning · Streaming · Structured Output · VisionFunction Calling · Reasoning · Streaming · Structured Output · VisionFunction Calling · Reasoning · Streaming · Structured Output · VisionFunction Calling · Reasoning · Streaming · Structured OutputFunction Calling · Reasoning · Streaming · Structured OutputStreaming · Structured Output · VisionFunction Calling · Streaming · Structured Output
Plan2,500 IQD / mo2,500 IQD / mo2,500 IQD / mo2,500 IQD / mo2,500 IQD / mo2,500 IQD / mo2,500 IQD / mo2,500 IQD / mo

Reasoning & agents

Qwen3.5 for the hardest problems, Qwen3.8 for everyday agent loops with tools.

Chat at volume

Qwen3.5-Lite and gpt-oss-lite keep latency and cost low for support, drafting and classification.

Documents & screenshots

Qwen2.5-VL reads receipts, dashboards and table images; Qwen3.8 is vision-capable too.

Accepting teams now

Bring your own editor.
We'll bring the models.

Tell us what you are building and we will provision a key, tune your limits, and stay on the line while you ship.

Get API accessConnect with us on WhatsAppor try the playground
# four lines, that is it
baseURL: "https://api.kalla.website/v1"