Is GPT-5.6 released yet?

Not as of June 12, 2026. OpenAI has been running internal evaluations under codenames 'ember-alpha' and 'beacon-alpha' since shortly after GPT-5.5's April 24, 2026 release. Multiple third-party trackers expect the release as early as June 2026. OpenAI has not officially confirmed a date. The S-1 filing on June 8 has likely slowed external communications during the SEC quiet period.

What features are expected in GPT-5.6?

Three leaked themes. First, context window expansion to 1.5M tokens (up from GPT-5.5's 1M), with internal MRCR v2 tests at the new range. Second, an UltraFast Codex mode targeted at competing with Claude Code on speed for coding workloads. Third, deeper agentic workflows — multi-step tool use with reduced supervision. The 'ember-alpha' codename has been associated with agentic features; 'beacon-alpha' with the long-context push.

Should I wait for GPT-5.6 or use GPT-5.5 now?

Use GPT-5.5 now. It is generally available in the API since April 24, 2026, scores 74.0% on MRCR v2 at 512K–1M tokens, and is priced at ~$5/$15 per million tokens — the cheapest frontier model. GPT-5.6 may or may not arrive in June, and pricing/availability will not be clear at launch. For production workloads that need stable contracts, stay on GPT-5.5 until GPT-5.6 has 30+ days of public benchmarks.

Will GPT-5.6 beat Claude Fable 5 on coding?

Unclear. Claude Fable 5 leads SWE-Bench Pro at 80.3% as of its June 9, 2026 release. GPT-5.5 scores 58.6% on the same benchmark. Closing a 22-point gap in one minor release would be exceptional. The 'UltraFast Codex' framing suggests OpenAI is targeting speed and cost-per-completion rather than raw SWE-Bench Pro accuracy. Realistic expectation: GPT-5.6 narrows the gap to 10–15 points, not full parity.

Quick Answer

GPT-5.6 Leaked Features: What to Expect From OpenAI's June 2026 Release

Published: June 12, 2026

GPT-5.6 Leaked Features: What to Expect From OpenAI’s June 2026 Release

Multiple third-party trackers expect OpenAI to ship GPT-5.6 in June 2026. OpenAI has been running internal evaluations under codenames “ember-alpha” and “beacon-alpha” since shortly after GPT-5.5’s April 24 release. Here is what is credible vs speculation.

Last verified: June 12, 2026

TL;DR

Element	Status
Expected release window	June 2026 (not yet confirmed)
Predecessor	GPT-5.5 (released April 24, 2026)
Leaked codenames	ember-alpha, beacon-alpha
Leaked context window	1.5M tokens (up from 1M)
Leaked feature	UltraFast Codex mode
Pricing	Unknown (likely at or near GPT-5.5 tiers)
Confidence	Medium-low on date, medium on context window, medium on Codex mode

What we actually know

OpenAI released GPT-5.5 on April 24, 2026 in the API. It also released GPT-5.5 Instant as ChatGPT’s new default model. Capability deltas vs GPT-5.4:

MRCR v2 at 512K–1M tokens: 74.0% (vs GPT-5.4’s 36.6% at the same range).
SWE-Bench Pro: 58.6% (improved over GPT-5.4).
Pricing: ~$5 in / $15 out per million tokens.

Since April 24, OpenAI has run a series of public-facing iterations on ChatGPT (GPT-5.5 Instant updates) and internal evaluations under the two codenames.

For the broader benchmark context, see Claude Fable 5 vs GPT-5.5 vs Gemini 3.5 Pro.

The three leaked themes

1. Context window expansion to 1.5M

The “beacon-alpha” codename has been associated with long-context internal testing. The leak: 1.5M token context window, up from GPT-5.5’s 1M. If real, this puts GPT-5.6 mid-way between GPT-5.5 (1M) and Gemini 3.5 Pro (2M).

The real question is not size but quality at the new range. MRCR v2 at 1M–1.5M will be the benchmark to watch.

2. UltraFast Codex mode

The “ember-alpha” leaks reference an “UltraFast Codex” mode. Likely framing: speed-optimized inline completions for IDE workloads, going head-to-head with Claude Code latency.

This makes strategic sense. Claude Code shipped agentic improvements with Claude Fable 5 on June 9, and Cursor 4 / Windsurf both default to non-OpenAI models for many tasks. UltraFast Codex would be OpenAI’s pitch to keep Codex CLI competitive.

See Claude Code vs Codex CLI vs Gemini CLI.

3. Deeper agentic workflows

GPT-5.5 added the Operator/Agent surfaces. GPT-5.6 leaks suggest deeper tool-use chains with reduced supervision and longer episode lengths. This is the response to Claude Fable 5’s long-horizon agent design.

Why the date is uncertain

OpenAI filed its S-1 on June 8, 2026. SEC quiet-period rules constrain marketing communications around a registration. OpenAI has not officially confirmed a GPT-5.6 release window since the S-1 filing. Possible scenarios:

Scenario A: GPT-5.6 ships in late June, framed as a technical update (not marketing). SEC quiet period permits product updates if framed as ordinary course of business.
Scenario B: Slip to July/August to clear initial SEC review.
Scenario C: Held until the S-1 goes effective.

Third-party trackers lean Scenario A. OpenAI has not commented.

For the S-1 context, see OpenAI S-1 vs Anthropic S-1: key differences.

What it likely will NOT include

Frontier-class SWE-Bench Pro parity with Claude Fable 5. Closing a 22-point gap in one minor release is implausible. Realistic: 10–15 point gap, not full parity.
Pricing cuts. OpenAI is reinvesting heavily in compute (Stargate buildout). Expect GPT-5.5 pricing tiers to carry over.
Open weights. OpenAI has not signaled any change in the closed-weights posture for frontier models.

How to plan around GPT-5.6

Production workloads — stay on GPT-5.5 until GPT-5.6 has 30 days of public benchmarks. The cost of a regression in a frontier minor release is higher than the upside of being early.

Coding workflows — keep Claude Fable 5 in your routing for agentic SWE tasks. Use GPT-5.5 for long-context retrieval and price. Re-evaluate when GPT-5.6 Codex mode is benchmarked.

Long-context RAG — wait for the 1.5M context window to be confirmed and measured before refactoring chunking strategies.

Decision matrix (June 12, 2026)

Workload	Today	Watch for GPT-5.6
Agentic coding	Claude Fable 5	UltraFast Codex mode
Long-context retrieval	GPT-5.5 (74% MRCR v2 at 1M)	1.5M context quality
Cheap frontier $/token	GPT-5.5	Pricing carry-over
Default ChatGPT	GPT-5.5 Instant	Auto-upgrade likely
Multimodal frontier	Gemini 3.5 Pro	TBD

Bottom line

GPT-5.6 is expected in June 2026 but not confirmed. The credible leaks are 1.5M context and an UltraFast Codex mode. The S-1 quiet period adds uncertainty. Until benchmarks land, GPT-5.5 stays the OpenAI default for production.

Sources: CometAPI tracker, Geeky-Gadgets, Releasebot, OpenAI release notes (June 2026).