Claude Sonnet 4.5 on streaming video
Claude Sonnet 4.5 · Anthropic · Multimodal · Closed API
Claude Sonnet 4.5 is Anthropic’s balanced flagship tier, released in September 2025 with a 200K-token context window (1M in beta) and strong agentic tool use alongside long-horizon coding ability. On Overshoot it is available as a passthrough model, reachable through the same stream publishing and OpenAI-compatible chat-completions request shape used across the API.
Claude Sonnet 4.5 is available through the Overshoot API as a passthrough model (claude-sonnet-4-6, 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 (1M beta)
- License
- Proprietary
- Released
- Sep 2025
- Inputs
- Text, images, video frames
- Overshoot availability
- API passthroughas of 2026-07-14
What Claude Sonnet 4.5 is good at
Sonnet 4.5 pairs strong vision understanding with Anthropic’s emphasis on agentic tool use, so it reads scenes, screenshots, and documents accurately while reliably chaining tool calls across multi-step tasks. It also carries Anthropic’s strongest coding performance in the Claude line, which matters for applications that mix visual input with code generation or debugging.
The 200K-token context window, extendable to 1M in beta, gives it room to hold long conversation history, tool outputs, and multiple sampled frames without truncating earlier turns in a long-running session.
- Accurate visual reasoning paired with reliable tool use
- Strong long-horizon coding and multi-step planning
- A 1M-token beta context for extended agentic sessions
Running Claude Sonnet 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 drawn from the 600-second retention window.
As a passthrough model, Claude Sonnet 4.5 reaches the API through the same OpenAI-compatible endpoint used across Overshoot: the platform manages stream lifecycle, frame selection, and SSE streaming transport, then forwards the resolved request to Anthropic. The current passthrough id on Overshoot is claude-sonnet-4-6, as listed by GET /v1beta/models.
Claude Sonnet 4.5 versus Haiku and Opus
Sonnet 4.5 sits between Claude Haiku 4.5, which answers faster for lighter workloads, and Claude Opus, which pushes further on the hardest reasoning tasks at higher latency. Sonnet 4.5 is the default choice for agentic pipelines that need both solid vision and dependable tool-calling without committing to Opus-level cost.
Frequently asked questions
Can Claude Sonnet 4.5 analyze live video?
Yes. A live WebRTC stream is queried by referencing a frame or a recent segment with an ovs:// URL inside a chat-completions request. Overshoot handles the stream and frame selection, then forwards the request to Anthropic for Claude Sonnet 4.5 to answer.
Is Claude Sonnet 4.5 open source?
No. Claude Sonnet 4.5 is closed and served solely through Anthropic’s API. Overshoot reaches it as a passthrough model, with keys managed exactly as they are for its hosted open-weight models.
How large is Claude Sonnet 4.5’s context window?
Sonnet 4.5 supports a 200K-token context window by default, with a 1M-token option available in beta for applications that need to carry very long conversation or tool history alongside sampled video frames.
What use cases fit Claude Sonnet 4.5 best?
Sonnet 4.5 fits agentic applications that combine visual understanding with multi-step tool use, such as an assistant that watches a live feed, calls external tools, and writes or debugs code in response to what it sees.