Loading Anthropic & OpenAI 2026...
Every model, product, and feature both labs have shipped — from ChatGPT and Claude 1 through GPT-5.5 and Claude Opus 4.8. Organized by category, then by date.
ChatGPT supports typed and real-time spoken conversations; OpenAI separately offers speech, transcription and realtime voice models through its API.
ChatGPT Images 2.0 (April 2026) added improved text rendering, multilingual support, and more advanced image generation and editing.
The macOS app introduced reading context from coding apps; Codex evolved into a broad cloud and CLI software-engineering agent.
OpenAI launched apps in ChatGPT and the Apps SDK in October 2025 — interactive third-party interfaces operate directly within conversations.
OpenAI has consolidated many agentic capabilities into the Responses API and Agents SDK.
By 2026, OpenAI positioned GPT-5.4 and GPT-5.5 as professional, tool-using models spanning ChatGPT, the API and Codex.
Free research preview using GPT-3.5
Premium subscription with priority access
Available in ChatGPT and Bing
Including Browsing and Code Interpreter
Users can exclude chats from training
Build tailored GPTs with skills and knowledge
Replaced by custom GPTs
Faster multimodal model — advanced capabilities reach free users
Web answers with source links
Workspace surface for writing and coding
Daily ship cadence starting Dec 5
Web search, file search/RAG, Code Interpreter, Computer Use via API
Solidified the OpenAI agent framework
Expanded context, lower cost
Reasoning with tool access
Advanced coding agent for Pro, Enterprise, Team
Flagship replacement for GPT-4o — coding, writing, health, reduced hallucinations
Most capable model — complex coding, research, data analysis across tools
Default model — reduced hallucinations, better personalization
New realtime voice models with reasoning + translation
Mobile/remote Codex access
For vetted developers and U.S. government partners
Claude began as Claude and Claude Instant in March 2023, followed by the public Claude.ai beta and API with Claude 2 in July 2023.
A defining Anthropic distinction — Claude creates persistent, editable output alongside the chat instead of plain text.
Launched April 2026 through Anthropic Labs for designs, prototypes, slides, and one-pagers.
Anthropic publishes Claude Code's functionality in a dedicated changelog — it evolves faster than any other product surface.
Anthropic created and open-sourced the Model Context Protocol — now a cross-lab ecosystem standard.
Anthropic has concentrated more on text, code, documents, and computer use than on native image/audio/video generation.
Anthropic maintains distinct release notes for the Claude consumer apps, API platform, system prompts, and Claude Code.
Claude for Small Business (May 2026) packaged connectors and ready-made workflows around common small-company tools.
Claude 3 introduced Haiku/Sonnet/Opus tiering; Claude 4 generations emphasized coding, agents, computer use, and long-running professional work.
First public model via the Anthropic API
First frontier model to reach it
Reduced hallucinations, better document handling
Opus / Sonnet / Haiku tiering with vision
Opus-level at Sonnet pricing — best coding model
Asana, Canva, Replit started building agents on it
Shipped alongside Claude 3.7 Sonnet
Claude Code hit $1B annualized run-rate in ~6 months
Grew to 200+ connectors
77.2% SWE-bench Verified, 30+ hour focus windows
Reclaimed the coding crown
Experimental arm co-led by Mike Krieger
Sandboxed VM — Claude reads/creates/edits files autonomously
Slack, Canva, Figma, Box, Clay
Super Bowl campaign as OpenAI tested ads
1M context, Agent Teams, 78.3% MRCR v2, 14.5-hour tasks, native PowerPoint
For Enterprise plans
60.4% ARC-AGI-2, 79.6% SWE-bench, 89.9% GPQA Diamond
IB, HR, PE, wealth, design; private plugin marketplaces; PPT/Excel export
Cross-device syncing
$100M investment in partner training/support/marketing
Fully autonomous multi-day task execution
AllTrails, Audible, Booking.com, Instacart, Spotify, Uber, Tripadvisor + more
Current frontier model
Parallel agentic execution across repos for enterprise early access
| Area | OpenAI | Anthropic |
|---|---|---|
| Consumer assistant | ChatGPT | Claude |
| Coding agent | Codex | Claude Code |
| Editable workspace | Canvas | Artifacts / Claude Design |
| Deep research | Deep Research | Research |
| Web access | Search + agent browser | Web Search, Web Fetch, Research |
| Computer control | Operator + Computer Use | Computer Use |
| Tool ecosystem | Apps SDK, Actions, connectors, MCP | MCP, connectors, Agent SDK |
| Image generation | Strong native product | No comparable flagship generator |
| Video generation | Sora | No comparable native product |
| Voice | Advanced Voice + Realtime API | Voice Mode (less central) |
| Enterprise knowledge | Company knowledge + connectors | Enterprise search + connectors + MCP |
| API architecture | Responses API + Agents SDK | Messages API + Agent SDK |
| Custom assistants | Custom GPTs | Projects, Styles, Artifacts, Skills, agents |
| Browser product | ChatGPT Atlas | Browser/computer use via Claude + agents |
| Design product | ChatGPT Images, Canvas, presentations | Claude Design + Artifacts |
Claude Opus 4.8, announced May 28, 2026, is the current frontier Anthropic model. Claude Sonnet 4.6 has been the default for Free and Pro users since February 17, 2026. Claude Opus 4.6 introduced the 1M-token context window and Agent Teams in February 2026.
GPT-5.5 (launched April 23, 2026) and GPT-5.5 Instant (May 5, 2026) are the current default models — built for complex coding, research, and data analysis across tools, with reduced hallucinations and improved personalization.
Both went from research previews to flagship coding agents within a year. Claude Code originated as a side project by Boris Cherny in late 2024, hit GA with Claude 4 in May 2025, and reached ~$1B annualized run-rate within six months. Codex (relaunched May 16, 2025) expanded into a cloud + CLI agent and went 'everywhere' (mobile/remote) on May 15, 2026. By June 2026 both labs treat coding agents as core platform surfaces.
Anthropic created and open-sourced the Model Context Protocol — an open standard for connecting LLMs to tools and data. OpenAI added native MCP support to the Agents SDK and the Responses API in 2025. By 2026 MCP is the cross-lab default for tool/data connectivity. Anthropic's connector directory exceeded 200 services in 2025; OpenAI's Apps SDK and connector list run in parallel.
Both ship weekly. OpenAI publishes separate release notes for ChatGPT, models, Business, Enterprise, Edu, desktop, and browser surfaces. Anthropic publishes distinct notes for Claude consumer apps, the API platform, system prompts, and Claude Code. Each lab now contains hundreds of individually trackable features.
OpenAI leads in consumer reach (ChatGPT), image generation, video (Sora), realtime voice, and platform breadth (Responses API, Agents SDK, Atlas browser, Apps SDK). Anthropic leads in agentic coding (Claude Code, Opus 4.x SWE-bench scores), MCP standardization, Artifacts and Claude Design as editable surfaces, and high-trust enterprise verticals (financial services, government, healthcare).
Anthropic Labs launched January 11, 2026 as an experimental product arm co-led by Mike Krieger (Instagram co-founder). It moves fast on consumer product ideas outside the core research roadmap. Early Labs ships include Claude Cowork (Jan 12, 2026) and Claude Design (April 2026).
No — on February 4, 2026, Anthropic publicly committed to keeping Claude ad-free, running a Super Bowl campaign while OpenAI announced it was testing ads (May 7, 2026 announcement).