Model capabilities

Capability matrix for Berget AI language models, including tool use, JSON output, streaming, and multimodal input

The table below lists the supported capabilities for each language model available on Berget AI. You can also verify capabilities against our API using our Test matrix.

Language models

ModelBasic chatTool useJSON modeJSON schemaStreamingMultimodalLong context + JSON
meta-llama/Llama-3.1-8B-Instruct
meta-llama/Llama-3.3-70B-Instruct
mistralai/Mistral-Small-3.2-24B-Instruct-2506
openai/gpt-oss-120b
zai-org/GLM-4.7-FP8

Capability definitions

Each column in the table above maps to one of the following capabilities.

CapabilityDescription
Basic chatStandard chat completion (/chat/completions).
Tool useFunction calling via the tools parameter.
JSON modeStructured JSON output via response_format: { type: "json_object" }.
JSON schemaStrict schema validation via response_format: { type: "json_schema", json_schema: ... }.
StreamingReal-time token streaming via server-sent events (stream: true).
MultimodalImage input in the content array (vision).
Long context + JSONJSON output on long-context requests (~8k output tokens).

On this page