Claude Sonnet 5.5 cuts agent costs
The build stack got cheaper, faster, and easier to govern today. Here are four changes worth knowing before your next model, voice, or agent-platform decision.
In today’s edition
- Claude Sonnet 5.5 gets faster at the same price.
- ElevenLabs launches low-latency expressive voice models.
- Base44 brings existing repositories into shared cloud workspaces.
- Perplexity adds reusable, versioned agent configurations.
Claude Sonnet 5.5 gets faster without raising API prices
Anthropic has released Claude Sonnet 5.5, positioning it as a faster, lower-cost model for everyday coding, bug fixes, documents, slides, and spreadsheets.
- API pricing stays at $2 per million input tokens, $10 per million output tokens, and $0.20 per million cache-read tokens.
- Anthropic says the model generates output 30%+ faster and costs up to 30% less per task than Sonnet 5 because it uses fewer tokens.
- On several of Anthropic’s evaluations, Sonnet 5.5 at Low or Medium effort beat Sonnet 5’s best score for about one-tenth of the cost per task.
- It is available across Claude’s platforms now, including AWS, Google Cloud, Microsoft Azure, and the Claude Platform as `claude-sonnet-5-5`.
What it means Move well-scoped coding and workflow jobs to Sonnet 5.5 first, where faster output and fewer tokens could cut both wait time and runtime spend.
Source: Anthropic
ElevenLabs targets natural voice agents at 100 ms latency
ElevenLabs has launched Eleven v4 and Eleven v4 Turbo, pairing a new expressive speech model with a low-latency version built for real-time conversations.
- Eleven v4 Turbo has roughly 100 ms median inference latency, according to ElevenLabs.
- The models support 90+ languages and let builders direct delivery with inline tags for emotion, pacing, reactions, sound effects, and style.
- Improved IPA support targets more reliable custom pronunciations, while the new architecture aims for more natural multi-speaker dialogue.
- Both models are available in ElevenAgents, ElevenCreative, and through ElevenAPI.
What it means Test v4 Turbo where your voice agent sounds flat or responds too slowly, since it combines real-time latency with controls for how each line is delivered.
Source: ElevenLabs
Base44 puts existing GitHub repos in shared cloud environments
Base44 has introduced Base Code, an early-preview workspace that scans an existing repository, assembles its development environment in the cloud, and gives the wider team a live place to work.
- Base Code is designed to set up repository dependencies such as databases, Redis, infrastructure components, and mock data in the cloud.
- Team members can work from a browser, see active sessions, jump into branches, open live previews, and review changes.
- The product aims to let product, design, QA, and other teammates contribute while engineering retains review and merge control.
- Base44 is offering the early preview free for 30 days.
What it means Use a shared cloud environment to bring non-engineering teammates into changes on a real codebase without making local setup and merge authority the bottleneck.
Source: MaorShlomo
Perplexity lets teams reuse versioned agents across apps
Perplexity’s Agent API now supports Profiles, Skills, and managed connectors, so a team can define an agent once instead of copying its configuration into every workflow.
- A Profile stores a model, instructions, tools, connectors, and run settings under one versioned ID.
- Skills are versioned procedures, letting teams update a workflow without editing every application that calls it.
- Project admins can configure connectors once, keeping individual apps from holding their own credentials.
- Profiles and Skills are live; managed connectors are in preview, starting with GitHub, Slack, Google Drive, Datadog, Linear, and Notion.
What it means Centralize shared agent instructions and approved access before your integrations multiply, so procedure updates and credential control do not become an app-by-app maintenance job.
Source: Perplexity Developers