github MCP (GitHub org) · 8h ago · mcp

MCP Python SDK v2.3.0 adds SSE event size limit and stricter header validation

from English

The Model Context Protocol (MCP) Python SDK has released version 2.3.0, introducing new configuration options alongside several behavioral changes and bug fixes.

Adds a `max_sse_event_size` parameter to `streamable_http_client` and `StreamableHttpParameters`, allowing users to configure the limit for Server-Sent Events (defaulting to 1 MiB).
Enforces stricter validation for tool annotations, causing registration to fail immediately if an `x-mcp-header` annotation is invalid.
Stops sending empty `_meta` and `params` fields in requests for connections prior to the 2025-11-25 protocol version, aligning with v1 behavior.
Omits the `experimental` capability from the `initialize` result when no experimental features are configured.
Pauses request timeouts during interactive OAuth login flows to prevent premature failures.

These updates improve compatibility with servers that reject empty metadata fields and provide better control over SSE event sizes for large tool results.

Importance 1/3 Trust 2/3 MCP (GitHub org) API & product launches

Read original

media MarkTechPost · 1h ago Live

Microsoft AI releases MAI-Transcribe-2-Streaming, ranking #1 on Artificial Analysis

Microsoft AI has released MAI-Transcribe-2-Streaming, its first streaming speech-to-text model, which ranks #1 of 38 models on the Artificial Analysis AA-WER Streaming index. Launched on October 1, 2026, the model transcribes 60 languages with continuous automatic language detection and targets voice agents and live captions.

github llama.cpp · 6h ago · 2 views

llama.cpp adds /v1/systemone API for laya, julia-1, lev, openjev, and kev models

The llama.cpp project has introduced a new `/v1/systemone` API endpoint to support specific model architectures including laya, julia-1, lev, openjev, and kev. This update includes the necessary server code, conversion scripts, and documentation to enable these models within the llama.cpp ecosystem.

lab Google — The Keyword (AI) · 12h ago · 6 views

Google announces Gemini 4 Argon and September AI updates

Google released its newest frontier model, Gemini 4 Argon, featuring advanced reasoning and a 1-million-token context window designed for complex challenges like cybersecurity defense. The company also announced several other products and scientific milestones from September, including the Gemini 3.8 Flash series, new voice models, and the AlphaGenome Atlas.

lab Anthropic News · 12h ago · 10 views

Anthropic launches $100M Claude Frontier Academy to train 10,000 engineers by 2027

Anthropic has launched the Claude Frontier Academy, a new training program backed by a $100 million commitment aimed at solving the enterprise AI talent gap. The initiative seeks to train 10,000 "Frontier Deployed Engineers" (FDEs) by the end of 2027, starting with cohorts from major firms like Accenture, McKinsey, and Morgan Stanley.

lab OpenAI News · 13h ago · 5 views

Chatham Financial uses OpenAI models to cut trade validation time from 30 minutes to under 4

Chatham Financial is integrating OpenAI's Codex and GPT-5.6 models into its capital markets workflows, significantly reducing the time required for routine tasks like trade validation.