Editorial illustration for Anthropic's Message Threads cut AI response time by 35% for long chats
Anthropic's Message Threads Cut AI Response Time 35%
Anthropic shipped a fix for one of the more wasteful quirks of API-based chat: resending an entire conversation history every single time a user sends a new message. The company's new feature, Message Threads, is now in public beta on the Claude platform, and it changes how long conversations get processed behind the scenes. Instead of transmitting the whole back-and-forth with each request, the system sends only what changed since the last message.
The payoff shows up in speed. For conversations stretching between 500,000 and one million tokens, Anthropic says time to first token drops by roughly 35 percent. That's the kind of gain that matters most to coding agents, computer-use tools, and any workflow chewing through large batches of documents or images, where long-running sessions are the norm rather than the exception.
Token usage itself stays the same, since Claude still has to read the full history to generate a response. What changes is how that history gets moved around. Threads stay encrypted in memory and expire after five minutes without activity. The feature is live on Claude's own platform as well as through AWS, Microsoft Foundry, and Google Cloud.
Message Threads only transmits what's changed since the last message, cutting time to first token by about 35 percent for requests between 500,000 and one million tokens, according to Anthropic.
Why this matters
For anyone running coding agents or long document-processing pipelines on Claude, resending the full conversation history on every call was a tax that got worse the longer a session ran. Message Threads cutting that overhead by 35 percent on 500K-to-1M-token requests is a real infrastructure fix, not a model upgrade dressed up as one. It won't make Claude smarter, but it should make agent loops feel less sluggish and probably trims API costs tied to redundant token transmission, though Anthropic hasn't put a number on that yet.
The beta tag matters here. Teams building production agents should test Message Threads against their own workloads before assuming the 35 percent figure holds outside Anthropic's benchmarks, since gains likely vary by how much context churns between turns. Still, this is the kind of unglamorous plumbing work that determines whether long-running agents are actually usable at scale. Watch whether competitors like OpenAI or Google ship something similar for their own long-context APIs, and whether Anthropic moves this out of beta with pricing changes attached.
Related Reading
Browse all Anthropic news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google to End Free Access to Gemini AI for Non-Subscribers
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic Claude gains shared context in Excel, PowerPoint; Microsoft adds Copilot Cowork
AI Tools & Apps
Related Topic Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source
From Archives OpenAI Details 2026 Election Stance as Chatbots Rely on Problematic Data
From Archives Apple's first robot proves harder than expected; AI robots still far from useful
From Archives Anthropic restricts Claude from recreating specific video game clones
AI Tools & Apps
From Archives Anthropic Posts USD 17B Q3 Revenue, Swings to Operating Loss on USD 5B Computing
Business & Startups
Common Questions Answered
How does Anthropic's Message Threads feature reduce API response time for long conversations?
Message Threads eliminates the inefficiency of resending the entire conversation history with each new message by only transmitting what has changed since the last message. This optimization cuts time to first token by approximately 35 percent for requests between 500,000 and one million tokens, significantly improving performance for extended chat sessions.
What is the specific performance improvement for Message Threads on large token requests?
According to Anthropic, Message Threads achieves a 35 percent reduction in time to first token for requests between 500,000 and one million tokens. This substantial improvement directly addresses the overhead that accumulates as conversation sessions grow longer and more token-intensive.
Why is Message Threads particularly beneficial for coding agents and document-processing pipelines?
Coding agents and long document-processing pipelines on Claude previously faced increasing overhead costs with each API call as they resent full conversation histories. Message Threads eliminates this tax by only transmitting incremental changes, making agent loops feel less sluggish and reducing API costs associated with redundant token transmission.
Is Message Threads a model upgrade or an infrastructure improvement?
Message Threads is an infrastructure fix rather than a model upgrade, meaning it does not make Claude smarter or improve its underlying capabilities. Instead, it optimizes how the API processes long conversations, resulting in faster response times and reduced operational costs without changing the model's actual performance or intelligence.
Further Reading
- Anthropic's Message Threads cut time to first token by 35 percent for long conversations - The Decoder
- Anthropic speeds up Claude Code with message threads. But your token bill stays the same. - The New Stack
- Anthropic's message threads let Claude agents skip resending chats - The Next Web
- Anthropic Announces Message Threads in Public Beta to Reduce Latency for Long-Running Sessions - Gadgets 360
- Is Claude Code truly faster with message threads? I did the math on my logs - ZDNET
