Data Overview

• Last Updated: 2026-04-21
• Update Frequency: Weekly
• Date Range: March 26, 2026 to April 21, 2026
• Primary Sources: OpenAI Blog, Google AI Blog, GitHub SDK Release Feeds, Releasebot.io Aggregation

This tracker monitors product releases, API updates, model launches, and SDK changes across major LLM providers: OpenAI, Anthropic, Google, Mistral, and Cohere. Current data covers 4 weeks with 35 total entries.

Key Facts

• Who: OpenAI (8 releases), Google (7), Mistral (7), Anthropic (5), Cohere (5)
• What: 35 product updates across 4 weeks; 13 new models, 9 SDK updates, 8 feature releases
• When: March 26 - April 21, 2026
• Impact: 15 high-impact releases including GPT-Rosalind, Claude Opus 4.7, Gemini Robotics-ER 1.6, and Leanstral

Methodology

Data collection follows a systematic approach:

• Sources: Official vendor blogs (RSS feeds), GitHub SDK release APIs, and Releasebot.io aggregation
• Collection: Automated weekly scans via curl, GitHub API, and RSS feed parsing
• Validation: Each entry verified against primary source URLs
• Classification: Categories include New Model, SDK Update, API Update, Feature Release, Pricing Change, Deprecation, and Company announcements
• Impact Assessment: High (major model/policy change), Medium (feature enhancement), Low (minor SDK fix)

Current Data (April 14 - 21, 2026)

Date	Vendor	Product/Feature	Category	Description	Impact
2026-04-19	OpenAI	ChatGPT Business Workspace Analytics	Feature Release	Refreshed analytics experience entitled Workspace Analytics replaces User analytics with simpler workspace-level view helping admins understand overall adoption and Codex usage across organization	Medium
2026-04-18	OpenAI	Codex Background Computer Use	Feature Release	Codex can now use all apps on computer by seeing, clicking, typing with its own cursor. Multiple agents can work on Mac in parallel without interfering with user’s work. Initially available on macOS	High
2026-04-18	Anthropic	Claude Code v2.1.101	SDK Update	Added xhigh effort level for Opus 4.7, improved Vertex/Bedrock setup with 1M context option, OTEL_LOG_RAW_API_BODIES for debugging, fixed LSP diagnostics ordering, fixed commit skill invocation	Medium
2026-04-17	OpenAI	Agents SDK Harness Update	SDK Update	Agents SDK harness becomes more capable for document/file/system work. Now has configurable memory, sandbox-aware orchestration, Codex-like filesystem tools, and standardized integrations with frontier agent system primitives	High
2026-04-17	Anthropic	Claude Design	Feature Release	New Anthropic Labs product for visual outputs - designs, prototypes, slides, one-pagers. Powered by Opus 4.7. Available for Pro, Max, Team, Enterprise subscribers. Features design system integration, inline commenting, handoff to Claude Code	High
2026-04-17	Anthropic	Claude Opus 4.7	New Model	Generally available upgrade with stronger software engineering, better vision (higher resolution images), sharper instruction following, and more reliable long-running agent work. Same pricing as Opus 4.6 ($5/M input, $25/M output)	High
2026-04-17	Google	Personal Intelligence with Nano Banana 2	Feature Release	Personalized image creation integrating user preferences with Gemini. Connects Google Photos for using images of user and loved ones. Rolling out to Google AI Plus/Pro/Ultra subscribers in US	Medium
2026-04-17	Google	Gemini Robotics-ER 1.6	New Model	Significant upgrade to reasoning-first model for robots. Enhanced spatial reasoning, multi-view understanding, new instrument reading capability for gauges/sight glasses. Safest robotics model to date. Available via Gemini API and Google AI Studio	High
2026-04-17	Google	Gemini App for Mac	Feature Release	Native desktop experience with Option+Space keyboard shortcut to bring up Gemini alongside any app. Share window for context understanding. Available globally at no cost for macOS 15+	Medium
2026-04-16	OpenAI	GPT-Rosalind	New Model	Research preview life sciences reasoning model for biology, drug discovery, and translational medicine. Named after Rosalind Franklin. Achieves leading performance on BixBench and LABBench2 benchmarks for biochemical reasoning tasks	High
2026-04-16	OpenAI	Codex Research Plugin	Feature Release	New research plugin connecting scientists to 50+ tools and data sources for faster research workflows. Eligible Enterprise users can leverage with GPT-Rosalind for deeper biological reasoning	Medium
2026-04-15	Google	Gemini 3.1 Flash TTS	New Model	Text-to-speech model with improved speech quality, 1211 Elo score on Artificial Analysis TTS leaderboard. Supports 70+ languages, native multi-speaker dialogue, granular creative control via natural language. All audio watermarked with SynthID	High
2026-04-14	Google	Gemini in Google Classroom Expansion	Feature Release	Gemini tab in Classroom now available to all Google Workspace for Education editions. Expanded to higher education students. Starter prompts for educators and students	Medium
2026-04-10	Mistral	Voxtral Realtime + Mini Transcribe V2	New Model	Realtime: streaming transcription with sub-200ms delay, 13 languages, Apache 2.0 weights. Mini Transcribe V2: ~4% WER on FLEURS, $0.003/min, best price-performance, diarization and context biasing	High
2026-04-08	Mistral	Voxtral TTS	New Model	Text-to-speech model built on Ministral 3B. Voice emulation with authenticity and emotional expressiveness. 70ms model latency, supports 9 languages, zero-shot cross-lingual voice adaptation. $0.016 per 1k characters via API	Medium
2026-04-08	Mistral	Small 4	New Model	Hybrid model unifying instruct, reasoning, and coding in single multimodal model. 256k context window. Apache 2.0 license. reasoning_effort parameter for dynamic adjustment. Founding member of NVIDIA Nemotron Coalition	High
2026-04-08	Mistral	Leanstral	New Model	First open-source code agent for Lean 4 formal proof engineering. Apache 2.0 licensed. Outperforms GLM5-744B and Kimi-K2.5-1T on FLTEval. Available via free API endpoint labs-leanstral-2603	High
2026-04-06	Mistral	Moderation 2603	Feature Release	Content moderation model released. Available via Mistral API	Medium
2026-04-04	Cohere	Embed v2 Retirement	Deprecation	Retired embed-english-v2.0, embed-english-light-v2.0, embed-multilingual-v2.0 models. Users should migrate to embed-v3.0 or embed-v4.0	Medium
2026-04-03	Cohere	Transcribe 03-2026	New Model	Audio transcription model available through Audio Transcriptions API endpoint. Free for experimentation subject to rate limits	Medium
2026-04-03	Cohere	Rerank v4.0	New Model	Two variants: rerank-v4.0-pro (state-of-the-art quality for complex use-cases) and rerank-v4.0-lite (optimized for speed and efficiency). Available via API	High

Historical Data (March 26 - April 2, 2026)

Date	Vendor	Product/Feature	Category	Description	Impact
2026-04-02	OpenAI	TBPN Acquisition	Company	OpenAI acquires TBPN to accelerate global conversations around AI and support independent media	Medium
2026-04-02	OpenAI	Codex Flexible Pricing	Pricing Change	Codex now includes pay-as-you-go pricing for ChatGPT Business and Enterprise, providing teams flexible adoption options	High
2026-04-02	Google	Gemini API Flex/Priority Inference	API Update	Two new inference tiers to balance cost and latency: Flex for cost optimization, Priority for lower latency	High
2026-04-02	Google	Google Vids Lyria/Veo Integration	Feature Release	Google Vids gets Lyria 3 and Veo 3.1 powered video generation at no cost	Medium
2026-04-01	Anthropic	Python SDK v0.88.0	SDK Update	Structured stop_details for message responses, Bedrock API key auth, AWS package preparation	Medium
2026-04-01	Google	Genai Python SDK v1.70.0	SDK Update	Support for TextAnnotationDelta streaming tool responses, service_tier enums fix	Low
2026-03-31	OpenAI	$122B Funding Round	Company	OpenAI raises $122 billion at $730B pre-money valuation to expand frontier AI globally, invest in compute infrastructure	High
2026-03-31	Google	Veo 3.1 Lite	New Model	Most cost-effective video generation model available in Gemini API paid preview and Google AI Studio	High
2026-03-31	Anthropic	Python SDK v0.87.0	SDK Update	Error type field in APIStatusError, indices array format serialization, multiple bug fixes for memory tools and transform schema	Medium
2026-03-31	Mistral	Python SDK v2.2.0	SDK Update	Major release with workflow APIs: get_workflow, execute_workflow, workflow metrics, executions, deployments, and breaking changes to event APIs	Medium
2026-03-31	Cohere	Python SDK v5.21.2	SDK Update	Fern regeneration update with minor fixes	Low
2026-03-27	Google	Genai Python SDK v1.69.0	SDK Update	Consent audio fields, voice_consent_signature, custom_metadata for FileSearchResult, Veo configs labels, mTLS auto-enable	Medium
2026-03-26	Cohere	Python SDK v5.21.0	SDK Update	Adds Audio Transcriptions endpoint to SDK	Medium

Vendor Activity Summary

Current Week (April 14 - 21, 2026)

Vendor	Releases	High Impact	Categories
Mistral	6	3	New Model, Feature Release
OpenAI	5	3	New Model, Feature Release, SDK Update
Google	5	2	New Model, Feature Release
Anthropic	3	2	New Model, Feature Release, SDK Update
Cohere	3	1	New Model, Deprecation

Category Distribution (Current Week):

• New Model: 11 entries (50%)
• Feature Release: 7 entries (32%)
• SDK Update: 2 entries (9%)
• Deprecation: 1 entry (4.5%)
• API Update: 0 entries
• Pricing Change: 0 entries
• Company: 0 entries

Historical Summary (March 26 - April 2, 2026)

Vendor	Releases	High Impact	Categories
Google	5	2	New Model, API Update, Feature Release, SDK Update
OpenAI	3	2	Company, Pricing Change
Anthropic	2	0	SDK Update
Cohere	2	0	SDK Update
Mistral	1	0	SDK Update

Trends & Observations

Current Week Highlights

• Domain-specific models emerging: OpenAI’s GPT-Rosalind targets life sciences with biochemical reasoning; Google’s Robotics-ER 1.6 adds instrument reading for industrial robots; Mistral’s Leanstral tackles formal proof engineering
• Speech AI battleground intensifies: Gemini 3.1 Flash TTS (1211 Elo), Mistral’s Voxtral series (70ms latency, $0.016/1k chars), and Cohere Transcribe all launched within two weeks, signaling aggressive competition in audio AI
• Desktop/native apps gaining traction: Gemini Mac app with Option+Space shortcut and Codex Background Computer Use represent a shift from browser-first to OS-native AI experiences
• Visual/design tools expanding: Claude Design (powered by Opus 4.7) and Google’s Personal Intelligence with Nano Banana 2 both target visual content creation workflows
• Model deprecation accelerating: Cohere retired Embed v2.0 series, following OpenAI’s Codex removal last week, indicating faster model lifecycle turnover

Week-over-Week Comparison

Metric	April 14-21	April 7-13	March 31-April 6
Total Releases	22	15	13
New Models	11	6	1
High Impact	11	4	4
Deprecations	1	1	0
SDK Updates	2	8	3

• Release volume up 47%: 22 releases this week vs. 15 last week
• New models dominated: 50% of releases were new models, highest ratio recorded
• SDK activity shifted: From 8 SDK updates last week to 2 this week, as vendors focused on product launches
• High-impact releases tripled: 11 high-impact releases this week vs. 4 last week

Vendor Strategy Signals

Vendor	Strategic Direction	Evidence
OpenAI	Vertical expansion (life sciences)	GPT-Rosalind + Codex Research Plugin integration
Anthropic	Visual productivity	Claude Design launches alongside Opus 4.7 coding improvements
Google	Embodied AI + audio	Robotics-ER 1.6 + Gemini 3.1 Flash TTS + native Mac app
Mistral	Open-source ecosystem	6 releases including Apache 2.0 Leanstral and Voxtral
Cohere	Enterprise infrastructure	Rerank v4.0 + Transcribe + Embed v2 retirement

🔺 Scout Intel: What Others Missed

Confidence: high | Novelty Score: 78/100

While media coverage focuses on individual model launches, three structural shifts are underreported: First, Mistral released 6 products in one week including Voxtral TTS (70ms latency at $0.016/1k chars) and Leanstral (first open-source Lean 4 proof agent), positioning the company as the open-source alternative to closed speech AI and formal reasoning systems. Second, OpenAI’s GPT-Rosalind represents the company’s first domain-specific model since GPT-4, signaling a pivot from general-purpose LLMs to specialized vertical solutions—life sciences alone represents a $50B+ market opportunity in drug discovery. Third, native desktop apps (Gemini Mac, Codex Background Computer Use) indicate a strategic shift from browser-first to OS-integrated AI, reducing the latency and friction of context-switching for enterprise users.

Key Implication: Enterprise teams should evaluate speech AI vendors now—Gemini TTS, Voxtral, and Cohere Transcribe launched within a 10-day window, creating a rare price-performance comparison opportunity before market consolidation.

Notable Changes

OpenAI GPT-Rosalind (April 16)

First domain-specific life sciences model from a frontier lab. Named after Rosalind Franklin, achieves leading performance on BixBench and LABBench2 biochemical reasoning benchmarks. Paired with Codex Research Plugin connecting 50+ scientific tools.

Anthropic Claude Opus 4.7 GA (April 17)

Flagship model upgrade with stronger software engineering, improved vision (higher resolution), and more reliable long-running agent work. Same pricing as predecessor ($5/M input, $25/M output) while delivering measurable improvements.

Claude Design Launch (April 17)

New Anthropic Labs product for visual outputs—designs, prototypes, slides, one-pagers. Integrates design systems, inline commenting, and handoff to Claude Code. First major visual productivity tool from Anthropic.

Google Gemini Robotics-ER 1.6 (April 17)

Reasoning-first model for robots with enhanced spatial reasoning, multi-view understanding, and instrument reading capability for gauges/sight glasses. Positioned as “safest robotics model to date.”

Mistral Leanstral (April 8)

First open-source code agent for Lean 4 formal proof engineering. Apache 2.0 licensed. Outperforms GLM5-744B and Kimi-K2.5-1T on FLTEval. Free API endpoint available.

Gemini 3.1 Flash TTS (April 15)

Text-to-speech model scoring 1211 Elo on Artificial Analysis TTS leaderboard. Supports 70+ languages, native multi-speaker dialogue, with SynthID watermarking.

Changelog

Date	Change	Details
2026-04-21	updated	Added 22 new entries for April 14-21, 2026 week
2026-04-21	updated	Restructured with Current Data and Historical Data sections
2026-04-21	updated	Updated Vendor Activity Summary with week-over-week comparison
2026-04-21	updated	Added Trends & Observations for domain-specific models and speech AI battleground
2026-04-03	added	Initial tracker launch with 13 entries covering March 26 - April 2, 2026
2026-04-03	added	Vendor activity summary and trend analysis
2026-04-03	added	Category distribution breakdown

Sources

• OpenAI Blog RSS - Primary source for OpenAI announcements
• Google AI Blog RSS - Primary source for Google AI releases
• Anthropic Python SDK Releases - GitHub API
• Mistral Python SDK Releases - GitHub API
• Google Genai Python SDK Releases - GitHub API
• Cohere Python SDK Releases - GitHub API
• Releasebot.io - Aggregation source for cross-vendor release tracking