Glimpse vs Jais 3

Glimpse

Glimpse reads images and video at $0.60 input and $3.00 output per million tokens, and OpenRouter serves it with tools, structured output and a 262,144-token context. It is a research preview with no public Elorian docs, model card or status page, and Elorian's terms allow training on submitted content unless a paid user opts out.

Jais 3

Jais 3 runs on the owner's hardware through OpenAI-compatible servers in vLLM and SGLang, with tool calls, three reasoning levels and a 512K context, from a 2.2 GB 0.9B model to a 758 GB 375B one. The weights download without sign-up under Apache-2.0, but every launch command trusts remote code, and no technical report or security contact is published.

Glimpse

1. Call it through OpenRouter as elorian/glimpse at https://openrouter.ai/api/v1/chat/completions. Elorian's own API docs sit behind a console sign-in
2. Send video as a video_url content part, as a public URL or a base64 data URL. Batch requests are not available for this model
3. Set max_tokens yourself. The endpoint allows up to 256,000 output tokens, and visual reasoning spends many of them
4. Honour Retry-After on 429 and 503. With a single provider, a 503 means the Together deployment is unavailable, not that another host can take over
5. Treat counts and timestamps as unverified. Elorian's terms say outputs need independent human review before anyone relies on them

Jais 3

1. Start vLLM or SGLang with --reasoning-parser k2_horizon, and add --tool-call-parser k2_horizon (vLLM also --enable-auto-tool-choice) for tools, or tool calls come back as raw text
2. Pass reasoning_effort (high, medium or low) and tool_call_format inside chat_template_kwargs. Other values return a template error
3. Send each earlier assistant turn back with reasoning_content as a string, even an empty one. A missing or null thinking field makes the template raise an error
4. Send your own system message, or set default_system_prompt to an empty string. Without one the template inserts an identity prompt naming IFM, Inception42 and Cerebras
5. Budget at least 32,768 output tokens at high effort, which the cards recommend for every request