Industry

OpenAI DevDay 2026 Complete Keynote Recap: GPT-6.1 Sol, Dots Coworkers, Ultrafast Mode, Agents API, and ChatGPT Space

Sam Altman presented the DevDay 2026 keynote at Fort Mason in San Francisco, unveiling OpenAI most comprehensive platform expansion to date. Key releases include GPT-6.1 Sol with an 80% price cut, always-on Dots coworkers, a 300 tok/s Ultrafast Mode, native Computer Use inside the Agents API, sub-second Decisions API, ChatGPT Space office suite, and a $500/mo Pro tier. Full executive breakdown and technical specifications.

By FreakVinci · 2026-09-29 · 22 min read

Fort Mason Keynote: OpenAI Shifts from Models to Autonomous Systems

On September 29, 2026, OpenAI held its annual developer conference, DevDay 2026, at the Festival Pavilion at Fort Mason Center in San Francisco. Delivered to an audience of three thousand engineers and streamed globally, CEO Sam Altman opened the address with a clear operational premise: rather than focusing solely on raw parameter scaling, the next frontier centers on latency, continuous autonomy, and economic efficiency.

Across ninety minutes of live demonstrations and technical reveals, OpenAI unveiled eight coordinated platform expansions spanning foundational models, inference infrastructure, developer APIs, and end-user collaborative suites.

Executive Summary: The Eight Major DevDay 2026 Announcements

OpenAI DevDay 2026 Platform Matrix
┌─────────────────────────────────────────────────────────────┐
│ 1. GPT-6.1 Sol (Production Frontier Model)                  │
│ • 96.8% Astra reasoning parity at 1/5th the inference cost  │
│ • $1.25 / 1M input | $5.00 / 1M output | 1M context window  │
├─────────────────────────────────────────────────────────────┤
│ 2. OpenAI Dots (Persistent Autonomous Coworkers)            │
│ • 24/7 background agent execution across Slack & GitHub     │
│ • Visual animated avatar interface with sandboxed microVMs  │
├─────────────────────────────────────────────────────────────┤
│ 3. Ultrafast Mode (High-Throughput Inference Tier)          │
│ • Sustained 280-320 tokens/second on NVIDIA Blackwell       │
│ • Sub-80ms TTFT powered by speculative multi-token decoding │
├─────────────────────────────────────────────────────────────┤
│ 4. Agents API with Native Computer Use                      │
│ • Standardized GUI actuation: cursor, clicking, keystrokes  │
│ • Managed Chromium browser sandboxes | 64.2% OSWorld 2.0    │
├─────────────────────────────────────────────────────────────┤
│ 5. Decisions API (Sub-Second Semantic Classifier)           │
│ • Sub-25ms classification and dynamic model routing         │
│ • Distilled Luna heads bypassing autoregressive generation  │
├─────────────────────────────────────────────────────────────┤
│ 6. ChatGPT Space (Collaborative Team Hub & Office Suite)    │
│ • Native Pages document authoring & automated Slides        │
│ • Direct competition with Google Workspace & M365 Copilot   │
├─────────────────────────────────────────────────────────────┤
│ 7. ChatGPT Pro $500/Month Subscription Tier                 │
│ • Unmetered deep research reasoning & priority GPU queues   │
│ • Concurrent execution of up to 10 autonomous Dots agents   │
├─────────────────────────────────────────────────────────────┤
│ 8. Realtime Voice & Vision API Updates                      │
│ • Sub-150ms speech-to-speech roundtrips with native pitch   │
│ • Direct camera video streaming over WebRTC channels        │
└─────────────────────────────────────────────────────────────┘

1. GPT-6.1 Sol: Resetting Frontier Economics

The centerpiece release was GPT-6.1 Sol (gpt-6.1-sol). Designed to address the unsustainable economics of deploying massive frontier reasoning models in high-throughput enterprise pipelines, Sol delivers:

  • Pricing: $1.25 per 1M input tokens, $5.00 per 1M output tokens (80% lower than GPT-6 Astra).
  • Context: 1-million-token native context buffer with 128k output token generation ceiling.
  • Verified Benchmark Parity: 74.8% on SWE-bench Verified (within 2.4 points of Astra), 64.2% on OSWorld 2.0, and 78.6% on GPQA Diamond.
  • Day-One Availability: Integrated immediately into the OpenAI API, Microsoft Azure AI Foundry, Amazon Bedrock, and GitHub Copilot.

2. OpenAI Dots: Always-On AI Coworkers

OpenAI moved beyond single-turn conversational chatbots with Dots, an autonomous agent platform featuring an expressive visual character avatar. Dots run continuously on managed cloud infrastructure:

  • Cross-Platform Integration: Connects via least-privilege OAuth to Slack, Linear, GitHub, and Google Workspace.
  • Overnight Execution: Dots process bug backlogs, review pending PRs, draft test suites, and generate morning briefing summaries while team members are away.
  • Security Sandboxes: Every agent session executes inside an isolated Firecracker microVM, requiring biometric human sign-off on irreversible production deployments or external communications.

3. Ultrafast Mode: Breaking the 300 Tokens/Second Ceiling

Addressing the execution speed bottlenecks that slow down autonomous multi-step loops, OpenAI previewed Ultrafast Mode:

  • Throughput: Sustained speeds between 280 and 320 tokens per second.
  • Mechanism: Combines a 3B-parameter distilled Luna draft model with parallel FP8 tensor verification on NVIDIA GB200 NVL72 racks.
  • Latency Impact: Compresses complex 1,000-token coding generations from 28.5 seconds down to 3.2 seconds.

4. Agents API with Native Computer Use

Expanding developer toolkits, OpenAI introduced direct operating system interaction inside the Agents API:

  • GUI Primitives: Translates model intentions into native mouse clicks, cursor drags, text typing, and shortcut keys.
  • Coordinate Normalization: Automatically maps a canonical 1024x768 visual space to host resolutions up to 4K.
  • Benchmark Performance: Reached 64.2% on OSWorld 2.0, surpassing competing desktop agents on complex multi-app tasks.

5. Decisions API: Sub-Second Classification

Targeting high-volume network gateways and ad bidding engines, the Decisions API delivers sub-25ms classification:

  • Non-Autoregressive Inference: Uses a single forward pass through a distilled Luna encoder to evaluate categorical schemas.
  • Ad Tech Integration: Partnered with Kevel to enable cookie-free real-time contextual ad matching within 50ms RTB bidding windows.
  • Cost: $0.05 per 1M input tokens and $0.01 per 10,000 classifications.

6. ChatGPT Space: The AI-Native Office Suite

OpenAI made its boldest move into enterprise productivity software with ChatGPT Space:

  • Pages: Real-time multi-user document authoring with embedded live code execution.
  • Slides: Automated presentation generation translating markdown outlines into editable 16:9 vector decks.
  • Agent Presence: Dots coworkers hold active seats in the workspace, drafting sections and auditing calculations alongside human team members.

7. ChatGPT Pro $500/Month Tier and Usage Adjustments

To balance cluster compute allocations, OpenAI structured a dedicated power-user tier:

  • $500/Month Pro: Unmetered deep research reasoning runs, dedicated Blackwell GPU queues, and up to 10 concurrent Dots coworkers.
  • $200/Month Tier Adjustments: Capped deep research at 50 sessions per day and background agents at 2 concurrent instances.

Strategic Impact: The Unified Autonomous Platform

OpenAI DevDay 2026 establishes a unified autonomous software stack. By pairing low-cost frontier intelligence (GPT-6.1 Sol) with high-speed execution (Ultrafast Mode), operating system actuation (Agents API), sub-second routing (Decisions API), and persistent background presence (Dots in ChatGPT Space), OpenAI has built an integrated operational environment for the next phase of enterprise software engineering.