Latest Feature Changelog
Live updates and newly introduced functionality in audited platforms. See what features been added or capabilities expanded.
ChatGPT
Spec logsGPT-6 Sol & Luna + Better Prompt Caching
GPT-6 Sol and GPT-6 Luna ship at 50% lower API prices than the GPT-6 promo rates; prompt caching gains a 30-minute prefix window with up to 90% discount on cached reads.
Claude
Spec logsClaude Opus 5.5 Release
Opus 5.5 matches Claude Fable 5.1 on most work at 40% lower run cost than Opus 5; $4/$20 per M tokens; external pre-release testing by Frontier Design and METR.
Mistral
Spec logsLe Chat Renamed Vibe
Mistral's official product FAQ confirms Le Chat is now named Vibe; the catalog entry was updated to match.
Claude
Spec logsAmodei: We Must Pace the Frontier
Anthropic commits to embedded third-party evaluators and proposes industry/global pacing steps after the OpenAI–Hugging Face agent incident.
Cline
Spec logsDesktop 0.0.25 — Codex Plan-Aware Model Picker
The ChatGPT Subscription (Codex) model picker now lists only the models your plan can actually use, removing non-Codex models and applying real backend context caps.
Suno
Spec logsSuno v6 — Music Industry Partnership Models
Announced a new generation of music models developed in partnership with the music industry, alongside global deals with BMG, Believe, and TuneCore.
Notion AI
Spec logsAdmin Model Controls for Agents
Workspace owners can now choose which AI models are available in Notion Agent and Custom Agents, set defaults, and pin model shortlists with speed/intelligence/cost scorecards.
vLLM Server
Spec logsv0.29.0 — Model Runner V2 Default
Model Runner V2 becomes the default for all models, with Mamba prefix caching gaining 9-25% time-to-first-token improvements and new model support (Hy4-preview, Qwen3.8-Flash-Next).
ChatGPT
Spec logsChatGPT Images 2.5
New image model with Sketch, comment-based editing, API models GPT-Image-2.5 Flare/Sunburst; up to 50% lower latency vs Images 2.0.
n8n
Spec logs2.39.0 — Atlassian 2LO Auth & OTel Tracing
Added Atlassian Service Account (2LO) auth for Jira/Confluence, Microsoft Teams Online Meeting resources, OTLP gRPC tracing support, and fixes for Anthropic Chat Model thinking output.
Mistral
Spec logs€3B Series D
Raised €3 billion in Series D funding at a post-money valuation above €21 billion to make sovereign, open-weight AI the technology frontier.
Claude Code
Spec logs2.1.265/266 — Plugin Dirs & Gateway Fix
Plugin folders via --plugin-dir with live add/remove, a 1GB cap on tool results saved to disk, prompt-cache-reuse fixes, and restoration of LLM-gateway/proxy sign-in behavior.
LM Studio
Spec logsBionic 1.1.2 — Linux Support & Accessibility
Added Linux builds for x64 and ARM64, improved screen-reader support, faster session loading, and llama.cpp 2.33.0 extension packs.
Ollama
Spec logsv0.34.0 Preview — Ollama Models in ChatGPT Desktop
Upcoming release lets local Ollama models be used inside ChatGPT Desktop, with faster structured outputs on Apple Silicon.
GitHub Copilot
Spec logsGPT-6 Astra Generally Available
OpenAI's GPT-6 Astra model reached general availability in GitHub Copilot, alongside continued weekly model lineup rotations (Gemini 3.8 Flash, Claude Fable 5.1).
Claude Code
Spec logs2.1.261 — Org Policy & Larger Tool Outputs
New Organization policy line in /status, bashOutputMaxChars/taskOutputMaxChars settings up to 128K chars, and --append-subagent-system-prompt-file for headless setups.
GitHub Copilot
Spec logsBusiness & Enterprise Signups Reopen
Copilot Business and Enterprise paid-tier signups reopened, enterprise-managed settings gained any-default-model support, and content exclusions reached GA in the Copilot app and CLI.
HuggingChat
Spec logsNVIDIA to Acquire Hugging Face
NVIDIA announced a ~$12.93B acquisition of Hugging Face; NVIDIA commits that HF stays an open, multi-cloud platform (close pending regulatory approval).
Claude Code
Spec logs2.1.260 — Diff Panel & Cache-Miss Diagnostics
Added a fullscreen diff panel beside the conversation (/diff), likely-cause reporting for prompt-cache misses in /cost, and multiple security fixes.
Cline
Spec logsAgent Plugins via Shared Hub
Agent Plugins are now discovered and run by the shared Hub, including automatic startup of their MCP servers.
Cursor
Spec logsSelf-Hosted Machines
Tool execution can run entirely inside your own network, with dynamic worker pools, existing sandbox integrations (AWS Lambda, Coder, Cloudflare, E2B), and computer-use support on Linux and Mac.
Ollama
Spec logsv0.33.3 — Multimodal Gemma 4 on MLX
gemma4 now supports images and audio on the MLX engine; cached prompt tokens are reported; GGUF default parameters are honored.
Open WebUI
Spec logsv0.11.2 — Security Fixes
Shipped security and access-control fixes; production deployments were advised to update promptly.
Open WebUI
Spec logsv0.11.3 — Accessibility Mode in Menus
Stronger background contrast now marks the menu entry under your pointer and the chosen model across dropdowns and the model picker in both themes.
Notion AI
Spec logsAgents Suggest Edits
Notion agents can now propose changes instead of applying them directly, with line-by-line user approval for granular passes like grammar fixes.
Ollama
Spec logsv0.33.2 — Dark Mode & Single-Instance Handoff
Desktop app follows system appearance again, hands off to an already-running instance on macOS, and keeps the Claude Desktop proxy stable during model catalog updates.
LM Studio
Spec logsBionic 1.1.0 — Agent Features & Vision Subagents
Shipped Auto Review for shell commands, vision subagents that let text-only models understand images, and 2-2.75x faster prompt processing on M5 Macs.
vLLM Server
Spec logsv0.28.0 — Kimi-K3 Performance Push
Major optimization effort for Kimi-K3 across the stack: Decode Context Parallel support, fused kernels, and ~17 GiB/GPU memory savings from optional shared-expert sharding.
ComfyUI
Spec logsv0.34.0 — HDR Video Saving
Added HDR video saving with AV1 codec plus mkv/webm, support for Pixal3d, TRELLIS2, and Sam3d-body models, and partner nodes for Wan 3.0 and Seedance 2.5 at 1080p.
Stable Audio
Spec logsSeries B — $232M Total Funding
Closed a Series B round backed by Electronic Arts, Sony Music Group, Universal Music Group, and Warner Music Group, bringing total funding to $232M under new leadership.
Open WebUI
Spec logsv0.11.1 — Streaming Rebuilt
Replies now stream as small deltas instead of resending the whole message each update, cutting data and server work on long replies by up to 1000x.
ElevenLabs
Spec logsElevenLabs CLI v1
Launched the first major version of the ElevenLabs command-line interface for developers working with audio tools in terminal workflows.
DeepSeek
Spec logsV4-Flash-Vision-Exp
Released an experimental vision variant of V4-Flash that additionally accepts image input; text capabilities match V4-Flash, with visual agent benchmarks bringing multimodal agent capabilities close to Opus-4.8 class.
Flux
Spec logsFLUX Upscale: 2K and 4K for Video
Shipped a dedicated FLUX Tool and API endpoint that regenerates any video at higher resolution, up to native 4K.
Mistral
Spec logsAgentic Search Launch
Shipped a retrieval layer helping AI systems navigate, read, and verify information inside and across complex documents.
Cursor
Spec logsCloud Agent Subscriptions & Subagents
Agents can now monitor PRs, watch Slack threads, or run scheduled tasks; adds custom modes, subagents on isolated VMs, and a long-lived /goal command.
Pika
Spec logsPika Speech, Soundtrack, Music & SFX
Shipped full-scene motion-aware soundtracks, text-to-music generation, real-time sound effects, and a 3B flow-matching text-to-speech model cloning voices from brief reference audio at 48kHz.
Stable Audio
Spec logsStable Audio 3.0 DAW Plugin
Introduced two new ways to work with Stable Audio 3.0: a DAW plugin and a more advanced generation experience on StableAudio.com.
Cursor
Spec logsOrigin Code Hosting (Early Beta)
Cursor now hosts code directly with repos, pull requests, code browsing, and two-way GitHub sync, alongside a Codebase tab and integrations with Vercel, Depot, and Buildkite.
Pika
Spec logsPika Audio Model Family
Debuted Pika's first family of foundation sound models — Soundtrack, Music, SFX, and Speech — available via the Pika API Club and billed as up to 20x cheaper than comparable audio models.
Suno
Spec logsStudio 2.0
Shipped the next version of Suno's browser-based DAW, rebuilt to let artists move at the speed of their ideas.
DeepSeek
Spec logsDeepSeek V4 Pro 0813 Update
DeepSeek V4 Pro line (GA Apr 2026) with the V4-Pro-0813 point release; V4 Flash (Jul 2026) leads open-weights Terminal-Bench.
Flux
Spec logsFLUX 3 Video: Generation
Published the first part of the FLUX 3 video capability series on the new Real World Models — multimodal flow models as the backbone of visual intelligence.
Google MusicLM
Spec logsSuperseded by Lyria / Flow Music
MusicLM was superseded by Lyria. Google retired MusicFX and MusicFX DJ on July 31, 2026, consolidating into Google Flow Music (launched April 18, 2026, powered by Lyria 3 / Lyria 3 Pro).
Claude
Spec logsClaude Opus 5 Implementation
Claude Opus 5 released with 1M context, $5/$25 pricing; Claude 3.7 retired Feb 19 2026.
Mini-SWE-Agent
Spec logsMini-SWE-Agent 2.4.6
Current 2.x series released; >74% on SWE-bench Verified; adopted by Meta, NVIDIA, IBM, Nebius, Anyscale, Princeton, and Stanford.
ChatGPT
Spec logsGPT-5.6 Family Launch
GPT-5.6 Luna/Terra/Sol released with reasoning tiers and 1.05M context window; GPT-4o retired Feb 13 2026.
Devin Desktop (formerly Windsurf)
Spec logsWindsurf Renamed Devin Desktop
Cognition relaunched Windsurf as Devin Desktop, an agent command center with a Kanban view over local and cloud agents. Devin Local replaced Cascade as the primary local agent; legacy Cascade remained usable through July 1, 2026. windsurf.com now redirects to devin.ai/desktop.
Codeium
Spec logsWindsurf Renamed Devin Desktop
After Cognition AI acquired Windsurf (July 14, 2025), the product is now named Devin Desktop (official FAQ: 'Devin Desktop is the new name for Windsurf').
GitHub Copilot
Spec logsUsage-Based Credits & Copilot Max
Introduced GitHub AI Credits (1 credit = $0.01) June 1, 2026. Added Copilot Max tier ($100/mo). New sign-ups for Pro/Pro+/Max paused (June 2026). Copilot coding agent and third-party agents (Claude Code, Codex) now accessible.
Replit Agent
Spec logsEffort-Based Pricing & Pro Tier
Agent v2 (2025) with up to 20-min autonomous runs. Effort-based pricing (Jun 2025). Teams plan sunset → Pro (Feb 2026). Starter free; Core $25/mo ($20 annual, $25 credits); Pro $100/mo ($95 annual, $100 credits, Turbo mode). Agent modes: Lite/Economy/Power.
Manus
Spec logsManus Agentic Expansion v1.0
Introduced advanced browser script execution helping speed up heavy research workloads.
Antigravity CLI
Spec logsAntigravity CLI Became Google's Terminal Experience
Google made Antigravity CLI available to everyone as the replacement terminal experience, and on June 18, 2026 Gemini CLI stopped serving requests for Google AI Pro, Google AI Ultra and free-tier individual accounts.
Gemini CLI
Spec logsGemini CLI Transitioned to Antigravity CLI
Google unified its agent-first developer tools into Antigravity, shipping Antigravity CLI as the replacement terminal experience. On June 18, 2026 Gemini CLI stopped serving requests for free-tier individual accounts, Google AI Pro, and Google AI Ultra users. Enterprise Gemini Code Assist licenses and API-key authentication are unaffected.
Grok
Spec logsGrok Vision Update
Introduced advanced image processing vision capabilities to premium users.
Gamma
Spec logsGamma Mobile Layout Upgrade
Optimized multi-grid slides to adjust screen layouts automatically for smartphones.
Udio
Spec logsUdio v1.5 Arranger
Introduced advanced segment arranger helping artists fine-tune individual bars.
Cursor
Spec logsAgentic Composer Terminal Diagnostics
Enables natural language task-runners to run compilers, look up logs, and resolve files on loop.
Perplexity AI
Spec logsPerplexity Pages Released
Enables research files to be exported into structured web articles with automatic layout graphics.
n8n
Spec logsn8n Advanced Agent Nodes v2
Launches visual vector database connectors and memory nodes for custom sub-agents.
AutoGen
Spec logsAutoGen Studio v2 Update
Introduced visual drag-and-drop editor to test and configure multi-agent networks within web apps.
Suno
Spec logsSuno v4 Launch
Released Suno v4 featuring 44.1kHz audio sampling and full multi-stem exports.
Mistral
Spec logsCodestral Release
Launched specialized coder weights cataloging efficiency across 80+ software frameworks.
Sora OpenAI
Spec logsSora Product Discontinued
OpenAI announced the discontinuation of the Sora video generation product effective April 26, 2026. Existing users were migrated to alternative video generation tools.
ElevenLabs
Spec logsVoice Cloning v2 Release
Introduced advanced emotional intensity tuning dials for cloned vocal speech.
Devin AI
Spec logsDevin Pricing Overhaul & 2.0 Launch
Devin 2.0 (agent-native IDE) announced April 2025. Pricing overhaul April 2026: old Core/Team ($500/mo) plans retired. Free $0; Pro $20/mo; Max $200/mo; Teams $80/mo base + $40/mo per full dev seat; Enterprise custom. Ask Devin/DeepWiki/Devin Review usage-billed.
CrewAI
Spec logsCrewAI Enterprise Studio
Launched visual drag-and-drop dashboard to build multi-agent crews without coding.
Zapier AI
Spec logsZapier Central Rollout
Launched an agent workspace enabling users to deploy autonomous workflow crawlers.
Llama
Spec logsLlama 4 Maverick
Llama 4 Maverick (128B, 1M context) remains Meta's flagship open-weights release (Apr 5 2025); no newer Llama generation has shipped.
Google Lyria 3
Spec logsLyria 3 Preview in Gemini API
Lyria 3 models (clip and pro) released 2026-03-25 in preview, global, 44.1 kHz / 192 kbps MP3 with SynthID + C2PA Content Credentials.
Notion AI
Spec logsGlobal Slack/Drive Integration
Allows Notion AI Q&A panels to find information stored in linked Slack and Drive folders.
Google Veo 3.1
Spec logsVeo 3.1 GA in Gemini API
Veo 3.1 reached GA as veo-3.1-generate-001. Veo 2 was deprecated and shut down June 30, 2026.
Gemini
Spec logsGemini 3.1 Pro Launch
Gemini 3.1 Pro/Flash released in preview with a 1M context window; Gemini 2.0 line superseded.
Qwen
Spec logsQwen3-Coder-Next Rollout
Alibaba releases Qwen3-Coder-Next (80B-A3B MoE) with strong SWE-bench Verified results, superseding Qwen2.5-Coder.
Phind Chat
Spec logsPhind Shut Down
Phind shut down January 16, 2026. phind.com returns 404; Pro subscriptions refunded and user data deleted by January 30, 2026. No successor or acquisition.
Phind-CodeLlama Pro
Spec logsPhind Shut Down
Phind shut down January 16, 2026, including Phind Pro. phind.com returns 404; Pro subscriptions refunded and user data deleted by January 30, 2026.
Udio
Spec logsDownloads Disabled Under UMG Partnership
Udio entered a settlement and partnership with Universal Music Group and disabled downloading of audio, video and stems. Creations are streamed inside a walled garden.
You.com Chat
Spec logsEnterprise API Pivot & Series C
You.com pivoted from consumer search to enterprise web-search APIs for AI (Web Search, Contents, Answer, Research, Finance + MCP server). Raised $100M Series C led by Cox Enterprises at a $1.5B valuation (Sept 2025).
OpenAI Operator
Spec logsOperator Integrated into ChatGPT
OpenAI fully integrated Operator into ChatGPT as agent mode; the standalone operator.chatgpt.com site sunset in the weeks after the official July 17, 2025 update.
Devin Desktop (formerly Windsurf)
Spec logsWindsurf Acquired by Cognition
Cognition (Devin's maker) acquired Windsurf on July 14, 2025.
MultiOn Web Agent
Spec logsMultiOn Consumer Agent Sunset
MultiOn sunset its consumer browser agent product in 2025. The company rebranded to AGI Inc. and pivoted to on-device mobile agents (AGI-0). The consumer browser agent product is no longer available.
Copilot Workspace
Spec logsTechnical Preview Concluded
GitHub concluded the Copilot Workspace technical preview on May 30, 2025. No new workspaces after that date; the GitHub Copilot Coding Agent (GA September 2025) is the successor.
Tome
Spec logsTome Shut Down
Tome's presentation app shut down April 2025 (data deleted; ~20M users). Founders pivoted to Lightfield (AI sales CRM). tome.app now returns 404; tome.com (Tome Law, AngelList-acquired) is a separate legal-AI product.
Codeium
Spec logsCodeium Rebranded to Windsurf
Codeium officially rebranded to Windsurf on April 4, 2025 (Codeium Extensions became Windsurf Plugins). The Codeium name is deprecated.
Midjourney
Spec logsMidjourney Version 7 Release
Introduced advanced photorealism and aligned spelling text layouts on generated assets.
Pika
Spec logsPika 2.0 Web Deployment
Rolled out improved physics modifiers allowing dynamic image squash/inflate animations.
Mutable AI
Spec logsMutable AI Acquired by Google
Mutable AI was acquired by Google in December 2024. The standalone product was discontinued and its technology integrated into Google's internal development tools. The Auto Wiki feature was relaunched as Google Code Wiki.
Stable Diffusion
Spec logsStable Diffusion 3.5 Large Release
Released upgraded weights with improved text spelling and prompt adhesion.
Flux
Spec logsFlux 1.1 Pro Launch
Released ultra-high speed generation API compiling prints 6x faster than previous scales.
Kling
Spec logsKling AI 1.5 Update
Released upgraded human kinetic rendering models with complete 1080p frame exports.
Ideogram
Spec logsIdeogram 2.0 Released
Rolled out upgraded model with 5x higher graphic design poster rendering accuracy.
Adept AI
Spec logsAdept Consumer Agent Discontinued
Adept's consumer-facing agent product was discontinued in June 2024. The company pivoted to an enterprise-focused model, licensing its technology to Amazon and continuing with a smaller team focused on enterprise solutions.
Runway
Spec logsRunway Gen-3 Alpha Turbo Implemented
Updated model compiling video frames 7x faster with full HD frame rates.
Drift AI Sales
Spec logsAcquired by Salesloft
Salesloft acquired Drift (announced Feb 13, 2024). drift.com now redirects to Salesloft's AI Chat Agent page ("We've transitioned from Drift to 1mind"). Clari+Salesloft announced a gradual Drift sunset in 2026, moving customers to 1mind.