Claude Haiku 4.5 on live video streams
Claude Haiku 4.5 · Anthropic · Multimodal · Closed API
Claude Haiku 4.5 is Anthropic’s fastest model, released in October 2025 with a 200K-token context window and near-frontier vision quality in a low-latency package. On Overshoot it is available as a passthrough model, reachable through the same stream publishing and OpenAI-compatible chat-completions request format used by every other model on the API.
Claude Haiku 4.5 is available through the Overshoot API as a passthrough model (claude-haiku-4-5-20251001, status: ready, verified against the live model catalog on 2026-07-14) - requests are forwarded to the upstream provider rather than running on Overshoot's hosted infrastructure.
- Developer
- Anthropic
- Parameters
- Undisclosed
- Context window
- 200K tokens
- License
- Proprietary
- Released
- Oct 2025
- Inputs
- Text, images, video frames
- Overshoot availability
- API passthroughas of 2026-07-14
What Claude Haiku 4.5 is good at
Haiku 4.5 packs quick first tokens and strong tool use into Anthropic’s smallest current model, closing most of the vision-quality gap with the larger Claude tiers while responding noticeably faster. It reads scenes, screenshots, and documents accurately and follows structured instructions well, which suits interactive or high-frequency visual workflows.
Because it is built for speed, Haiku 4.5 fits agentic loops that call a vision model repeatedly, such as monitoring a feed and reacting to changes, where waiting on a slower frontier model each turn would add up.
- Near-frontier visual quality with quick first-token latency
- Strong tool use for agentic, multi-turn workflows
- Reliable structured output for repeated, high-frequency queries
Running Claude Haiku 4.5 on Overshoot
Publish a camera or screen share over WebRTC through LiveKit to open a Stream, then send a chat-completions request whose image_url or video_url is an ovs:// reference to the latest frame, an exact timestamp, or a recent segment within the 600-second retention window.
As a passthrough model, Claude Haiku 4.5 is reached through the same OpenAI-compatible endpoint used across Overshoot: the platform handles stream lifecycle, frame selection, and SSE streaming transport, then forwards the resolved request to Anthropic. The current passthrough id on Overshoot is claude-haiku-4-5-20251001, as listed by GET /v1beta/models.
Claude Haiku 4.5 versus Claude Sonnet 4.5 and Opus
Haiku 4.5 is the tier to reach for when speed and cost matter more than maximum reasoning depth, while Claude Sonnet 4.5 trades some latency for stronger long-horizon coding and agentic performance, and Claude Opus targets the hardest reasoning tasks where latency matters least. All three share the same request shape on Overshoot, so switching tiers is a matter of changing the model id: for Haiku 4.5, the current passthrough id is claude-haiku-4-5-20251001.
Frequently asked questions
Can Claude Haiku 4.5 analyze live video?
Yes. Publish a live WebRTC stream, point a chat-completions request at the latest frame or a recent segment with an ovs:// URL, and Overshoot resolves that reference before forwarding the request to Anthropic for Claude Haiku 4.5 to answer.
Is Claude Haiku 4.5 open source?
No. Claude Haiku 4.5 is proprietary and reachable only through Anthropic’s own API. Overshoot exposes it as a passthrough option, handling keys the same way it does for the open-weight models it hosts.
How fast is Claude Haiku 4.5?
Haiku 4.5 is Anthropic’s fastest current model, built for quick first tokens. As a passthrough model on Overshoot, overall response time depends on Anthropic’s serving stack, but Haiku 4.5 is consistently the quickest option within the Claude family.
When should I choose Claude Haiku 4.5 over Sonnet or Opus?
Choose Haiku 4.5 for latency-sensitive or high-volume visual queries where near-frontier quality is enough. Move to Sonnet 4.5 or Opus when a task needs deeper reasoning or longer-horizon planning and can tolerate a slower response.