# GPT-5.5 Spreads Across AI Products as Agents Get a Real-World Test

*By AI News Digest • April 25, 2026*

The biggest story was downstream adoption: Microsoft, GitHub, Devin, Perplexity, OpenRouter, and Cursor all moved quickly on GPT-5.5. Anthropic added a more grounded agent story with a live negotiation experiment, while Cohere, Meta, and ComfyUI signaled larger shifts in sovereignty, compute, and tooling.

## GPT-5.5 becomes a distribution story

### Microsoft and developer tools move quickly
OpenAI's new model moved into major work products quickly. Microsoft said GPT-5.5 is rolling out to GitHub Copilot, M365 Copilot, Copilot Studio, and Foundry, where it is positioned for deeper reasoning, stronger multistep execution, and better performance on long, complex tasks; in Copilot CLI, users can switch models by job, while the Rubber Duck agent adds a multi-model review loop [^1][^2][^3][^4]. GitHub said GPT-5.5 is generally available and rolling out in Copilot, with early testing showing its strongest performance on complex agentic coding tasks and real-world coding challenges previous GPT models could not resolve; Cursor also said the model is now available there and currently leads CursorBench at 72.8% [^5][^6][^7].

**Why it matters:** The notable development today is how quickly GPT-5.5 is being embedded into everyday coding and enterprise workflows.

### Agent products adopt it as an execution layer
Cognition released GPT-5.5 in Devin as an Agent Preview, saying it runs longer and more autonomously than any GPT model it tested, surfaces bugs other models miss, and can investigate and fix production issues end-to-end [^8]. Perplexity is also rolling out GPT-5.5 as the default orchestrator model for Perplexity Computer, replacing Opus 4.7 as it monitors user sentiment during the rollout [^9]. OpenRouter said GPT-5.5 and GPT-5.5 Pro are live, describing GPT-5.5 as state of the art for long-running work across code, data, and tools, with Pro aimed at more complex reasoning and analysis [^10][^11].

**Why it matters:** This extends the story from a model launch to adoption inside products built for longer-running agent work.

## Agents get a more concrete market test

### Anthropic's negotiation experiment found demand and a hidden model gap
Anthropic said Claude agents interviewed 69 colleagues about what they wanted to buy and sell, then completed 186 deals worth more than $4,000; survey respondents generally viewed the outcomes as fair, and nearly half said they might pay for a service like this [^12][^13]. The company also found that model quality mattered materially: Opus got substantially better deals than Haiku in simulated runs, while participants did not notice the gap, and Anthropic says AI-agent markets may create value but still have rough edges that policy and legal frameworks will need to address [^14][^15]. It also logged some odd behavior, including one agent buying 19 ping-pong balls for itself and another buying a duplicate snowboard after inferring its user's taste from a casual mention of skiing [^16][^17].

**Why it matters:** This is a useful step beyond benchmark talk. It shows agents can transact in a small market, but it also shows that hidden model advantages can shape outcomes without users noticing.

## Strategic moves beyond the model race

### Cohere and Aleph Alpha pair up around sovereign AI
Cohere and Aleph Alpha said they are forming a transatlantic AI partnership anchored in Canada and Germany, combining Cohere's global scale with Aleph Alpha's European R&D to build sovereign, enterprise-grade AI with security, privacy, and trust as the focus [^18]. The announcement included executives from Cohere, Aleph Alpha, and Schwarz Digits alongside ministers from Canada and Germany, and Aidan Gomez framed the deal around deep Canada-Germany strategic backing and Germany's role as Europe's economic powerhouse [^18][^19].

**Why it matters:** This is a clear sign that sovereign AI is becoming a concrete strategy for competing for business and government demand.

### Meta adds AWS compute to its AI portfolio
Meta said it has agreed with AWS to bring tens of millions of AWS Graviton cores into its compute portfolio, expanding the infrastructure behind Meta AI and its agentic experiences that serve billions of people [^20].

**Why it matters:** As AI products become more agentic, infrastructure scale and supply diversification are becoming strategic differentiators.

### ComfyUI raises to scale open creative tooling
ComfyUI said it raised $30 million at a $500 million valuation, bringing total funding to $47 million, and reported 4 million users, more than 60,000 community-built nodes, and more than 150,000 daily downloads [^21]. The company said the funding will go toward Comfy Cloud, collaborative workflows, a better local experience, more dependable node infrastructure, and day-one compatibility for major model releases, while emphasizing open infrastructure rather than a walled garden [^21].

**Why it matters:** This is a notable funding signal for the open tooling layer that sits between fast-moving model releases and production creative workflows.

---

### Sources

[^1]: [𝕏 post by @satyanadella](https://x.com/satyanadella/status/2047743651053556126)
[^2]: [𝕏 post by @satyanadella](https://x.com/satyanadella/status/2047743653045837988)
[^3]: [𝕏 post by @satyanadella](https://x.com/satyanadella/status/2047743655587504507)
[^4]: [𝕏 post by @satyanadella](https://x.com/satyanadella/status/2047743658527789565)
[^5]: [𝕏 post by @github](https://x.com/github/status/2047747243617460482)
[^6]: [𝕏 post by @cursor_ai](https://x.com/cursor_ai/status/2047744579127185843)
[^7]: [𝕏 post by @gdb](https://x.com/gdb/status/2047753216264396877)
[^8]: [𝕏 post by @cognition](https://x.com/cognition/status/2047743153461936257)
[^9]: [𝕏 post by @AravSrinivas](https://x.com/AravSrinivas/status/2047753276888854595)
[^10]: [𝕏 post by @OpenRouter](https://x.com/OpenRouter/status/2047744317415141787)
[^11]: [𝕏 post by @gdb](https://x.com/gdb/status/2047755043278094837)
[^12]: [𝕏 post by @AnthropicAI](https://x.com/AnthropicAI/status/2047728365092663340)
[^13]: [𝕏 post by @AnthropicAI](https://x.com/AnthropicAI/status/2047728369169518678)
[^14]: [𝕏 post by @AnthropicAI](https://x.com/AnthropicAI/status/2047728371962888371)
[^15]: [𝕏 post by @AnthropicAI](https://x.com/AnthropicAI/status/2047728384503939464)
[^16]: [𝕏 post by @AnthropicAI](https://x.com/AnthropicAI/status/2047728378275381427)
[^17]: [𝕏 post by @AnthropicAI](https://x.com/AnthropicAI/status/2047728380997455928)
[^18]: [𝕏 post by @cohere](https://x.com/cohere/status/2047621404950942117)
[^19]: [𝕏 post by @aidangomez](https://x.com/aidangomez/status/2047629675468365876)
[^20]: [𝕏 post by @AIatMeta](https://x.com/AIatMeta/status/2047647617681957207)
[^21]: [𝕏 post by @ComfyUI](https://x.com/ComfyUI/status/2047731693788995871)