Breaking

OpenAI Ships Eight Models in Six Months, Previews GPT-5.6 Sol

June 29, 2026 at 08:45 EDT

Foundation Models
Media Generation
AI Agents

OpenAI released eight new models in roughly six months from December 2025 to June 2026, spanning flagship, compact, voice and image families, and has now opened a limited preview of its latest model, GPT-5.6 Sol. The lineup announced over the period comprises GPT-5.5, GPT-5.4 mini, GPT-5.4 nano, GPT-Realtime-2, GPT-Realtime-Whisper, GPT-Realtime-Translate, GPT-Image-2 and GPT-5.6 Sol—all confirmed as real releases via official announcements and API documentation.

Dec 2025 – Jun 2026 · OpenAI

Eight Models in Seven Months — and One Held Back

OpenAI shipped eight new models spanning flagship and compact GPT-5 variants plus dedicated voice and image systems — while its newest "GPT-5.6 Sol" stays in limited preview at the request of the U.S. government.

new models released in 7 months

~1.05M

token context window on GPT-5.5

700k+

GPU hours of red teaming on Sol

Terminal-Bench 2.1 — state of the art

GPT-5.6 Sol pushes the top-mode score on complex command-line workflows.

88.8

previous
best

91.9

GPT-5.6 Sol
(top mode)

The lineup at a glance

Price per 1M tokens · context · focus

GPT-5.5 $5 in / $30 out · ~1.05M ctx

Primary coding, research and data analysis

GPT-5.4 mini $0.75 in / $4.50 out · 400k ctx

2×+ faster than the larger model

GPT-Realtime-2 audio $32 / $64 · 128k ctx

Speech-to-speech with tool use

GPT-5.6 Sol $5 in / $30 out · limited preview

Cyber and agent enhancements · family of Sol, Terra, Luna

Held in limited preview

GPT-5.6 Sol is offered only to trusted partners following a U.S. government request — though OpenAI signals such restrictions "shouldn't be the norm."

Previewed June 26, 2026 — flagship plus Terra and Luna
More than 700,000 GPU hours of red teaming
State-of-the-art coding and cybersecurity workflows

What users praise

GPT-5.5: faster tasks, better token efficiency, strong Codex quotas
5.4 mini: low cost and longer usable sessions
Image-2: better photorealism and text accuracy

What draws criticism

GPT-5.5: practical gain, not a dramatic benchmark leap
Realtime-2: smarter but needs redesigned prompts
Sol: top-tier yet a gap to rivals on general benchmarks; limited release frustrates

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.