GPT-5.5: According to Alexandr Wang, Meta’s next frontier model, codenamed Watermelon, has already caught up with OpenAI’s GPT-5.5 on internal benchmarks. The model is still training.

wpnews.pro

cd /news/large-language-models/gpt-5-5-according-to-alexandr-wang-m… · home › topics › large-language-models › article

[ARTICLE · art-48541] src=outyet.ai ↗ pub=2026-07-03T00:00Z topic=large-language-models verified=true sentiment=· neutral

GPT-5.5: According to Alexandr Wang, Meta’s next frontier model, codenamed Watermelon, has already caught up with OpenAI’s GPT-5.5 on internal benchmarks. The model is still training.

OpenAI has released GPT-5.5, confirmed via official access on April 23, 2026. The model achieves a score of 96.6 on the Agentic#2 benchmark and 1474.62 on LM Arena, while Meta's rival model Watermelon reportedly matches its performance internally.

read1 min views1 publishedJul 3, 2026

GPT-5.5: According to Alexandr Wang, Meta’s next frontier model, codenamed Watermelon, has already caught up with OpenAI’s GPT-5.5 on internal benchmarks. The model is still training. — Image: Outyet (auto-discovered)

OpenAIGPT-5.5 Yes.

How we confirmed it

Released via official access

April 23, 2026 at 12:00 AM UTC

Waiting for OpenAI's next model?

One click creates your free account (or signs you into an existing one) and turns on alerts for OpenAI releases. There's no password to set, and you can unsubscribe any time.

Benchmarks

Agentic#2

GPT-5.5 · via BenchLM.ai · captured Jul 3, 2026 · source 96.6

FrontierCode Diamond

GPT-5.5 · via Epoch AI · captured Jul 3, 2026 · source 6.3

GDP.pdf

GPT-5.5 (xHigh reasoning) · via Epoch AI · captured Jul 3, 2026 · source 25

LM Arena#16

gpt-5.5 · via LM Arena · 38,470 votes · captured Jul 3, 2026 · source 1474.62

Sourced metrics mirrored with attribution — not community sentiment.

Data from [BenchLM.ai](https://benchlm.ai).

Data from [Epoch AI, “AI Benchmarking Hub”](https://epoch.ai/benchmarks) (CC BY 4.0).

Vibe rating

How this model actually feels to use, rated by the people here.

No vibe ratings yet. Be the first.

How is this model behaving today? One check-in per person per day: the running record of “did they nerf it”. Resets at midnight UTC.

No check-ins yet today. Be the first to log how this model is behaving.

An AI summary of the public comments here. Not a sourced fact.

Not enough public feedback yet to summarize. Add a comment to help build the picture.

[Compare GPT-5.5 with another model →](/compare?models=gpt-5-5)

Also waiting on: [ChatGPT 5.6](/models/chatgpt-5-6), [Gemini 3.5 Pro](/models/gemini-3-5-pro), [Claude Fable 5.1](/models/claude-fable-5-1), [Claude Mythos 5.1](/models/claude-mythos-5-1)

Comments #

Missing a model, or wishing OutYet did something it doesn't? Tell us what to build.

source & further reading

outyet.ai — original article Leanstral 1.5: ICYMI: Mistral released Leanstral 1.5, a SOTA open model for Lean 4 proof engineering. > Developers use Lean 4 as a general-purpose functional language (for CLI tools and libraries)… GPT-5.6 Sol: First reports about GPT-5.6; The interesting thing about GPT-5.6 Sol is not that it looks cleaner than Opus. It apparently doesn’t. Muse Spark: Meta is still training their big models but we will get an update on Muse Spark soon https://t.co/M62UrzO40q

~/api · this article 200

$curl api.wpnews.pro/v1/news/gpt-5-5-according-to-ale…

Read original on outyet.ai → outyet.ai/models/gpt-5-5

mentioned entities

OpenAI

GPT-5.5

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main

→ Live at https://your-agent.zahid.host ✓

Get free account → Pricing

from €0/mo · no card required