A leak on X with 2,890 likes and 207 retweets reveals key information about OpenAI’s next flagship model: GPT-6, internal codename “Goblin,” scheduled for September 29, 2026.
Leaked Information Summary
| Item | Content |
|---|---|
| Model Name | GPT-6 |
| Internal Codename | Goblin |
| Release Date | September 29, 2026 |
| Source | X platform leak post (2,890 likes, 380K views) |
| Credibility | Medium (not officially confirmed by OpenAI) |
Timeline: 5-Month Iteration Cycles
| Model | Release Date | Gap from Previous |
|---|---|---|
| GPT-5.4 | December 2025 | — |
| GPT-5.5 | April 23, 2026 | ~5 months |
| GPT-6 | September 29, 2026 | ~5 months |
Two consecutive 5-month iteration cycles — significantly faster than OpenAI’s previous flagship pace. GPT-4 to GPT-4o was ~10 months; GPT-4o to GPT-5 was ~14 months.
What “Goblin” Means
OpenAI’s model codenames are unconventional:
- GPT-5’s pre-training architecture was codenamed “Spud”
- GPT-5.5 continued Spud improvements
- “Goblin” suggests GPT-6 may adopt a completely new architecture rather than incremental upgrades
Combined with recent OpenAI moves:
- Ending Microsoft exclusivity: GPT-5.5 now on AWS Bedrock
- DeployCo JV: $500M joint venture with TPG, Bain for enterprise AI
- Codex Agent as standalone product: Coding Agent operates independently
The picture: OpenAI is shifting from “model as product” to “platform as ecosystem,” and GPT-6 may be the carrier of this transformation.
Competitive Landscape Projection
| Model | Expected Timeline | Positioning |
|---|---|---|
| GPT-6 (Goblin) | 2026.9 | OpenAI flagship, new architecture |
| Claude Opus 5.0 | 2026 Q4? | Anthropic flagship |
| Claude Sonnet 4.8 | 2026.5? | Mid-tier (confirmed in source leak) |
| DeepSeek V5? | 2026 H2? | Chinese open-source flagship |
| Meta Avocado | 2026.5 (delayed) | Meta next-gen model |
Signal vs Noise Assessment
While a single leak has limited credibility, these cross-references increase confidence:
- May roadmap leaks: Multiple sources mention密集 releases in May 2026, September flagship aligns with this rhythm
- OpenAI acceleration signals: Ending exclusivity, DeployCo, Codex independence all point to a larger product rhythm
- Competitive pressure: DeepSeek V4, Claude Opus 4.7, Qwen 3.6 are narrowing gaps — OpenAI has incentive to accelerate
Action Advice
| Your Situation | Recommendation |
|---|---|
| Evaluating long-term model vendor | Wait for GPT-6 architecture details before major decisions |
| Deeply invested in GPT-5.5 | 5.5’s lifecycle may be short (5 months), monitor migration cost |
| Open-source model users | The window before September is golden for DeepSeek V4/Qwen 3.6 habit formation |
| AI industry observers | Watch for June-July warm-up info, GPT-6 technical details may leak gradually |