GPT-5.4 Thinking and GPT-5.4 Pro are rolling out now in ChatGPT. GPT-5.4 is also now available in the API and Codex. GPT-5.4 brings our advances in reasoning, coding, and agentic workflows into one frontier model.
Image from OpenAI (@OpenAI)

Mar 5, 2026 · 6:10 PM UTC

2,033
3,192
1,988
23,110
7,207,937
RelevantRecentLikes
GPT-5.4 is our most factual and efficient model: fewer tokens, faster speed. In ChatGPT, GPT-5.4 Thinking has improved deep web research, better context retention when it thinks for longer—and oh—you can now interrupt the model and add instructions or adjust its direction mid-response. Steering is available this week on Android and web. iOS coming soon.
144
202
88
2,813
483,178
GPT-5.4 Thinking and Pro are rolling out gradually starting today across ChatGPT, the API, and Codex. openai.com/index/introducing…
Introducing GPT-5.4

Introducing GPT-5.4

Introducing GPT-5.4, OpenAI’s most most capable and efficient frontier model for professional work, with state-of-the-art coding, computer use, tool search, and 1M-token context.

openai.com
70
109
30
1,702
292,863
@OpenAI
Image from Damian Player (@damianplayer)
13
56
7
1,223
64,985
@OpenAI
Image from polyfield (@Polyfielder)
5
19
3
888
48,954
I mean
Image from Nathaniel W. (@natemixing)
23
10
395
39,584
Does this update include a fix to stop lying to your customers?
5
1
482
19,654
Thats great and all but…
Image from Zukao (@zukaodev)
6
5
345
15,124
@OpenAI
2
1
121
7,648
Progress is clearly accelerating. Things are going to get crazy.
Image from Dr Singularity (@Dr_Singularity)
1
8
83
6,501
Lets go! I added GPT-5.4 to our arena, it's already battling against Minimax 2.5 (previous winner) in Civilization and playing other games with other models. The model seems way smarter out of the gate. Excited to see how it fairs in long horizon, bluffing, and collaborative games
5
1
83
26,487
xitter.twitterapi.xyz/ValsAI/status/20296209…
Vals AI (@ValsAI) profile pictureVals AI
@ValsAIMar 5
GPT 5.4 is #1 on Vibe Code Bench at 67.4%, +5.7% higher than the previous SOTA. This is our benchmark that measures model’s ability to produce an entire working application from a short text specification.
Image from Vals AI (@ValsAI)
3
67
13,295
wish me luck lads
Image from 💺 (@patience_cave)
4
69
2,111
We’ve been testing GPT-5.4 all week at @every inside our @OpenClaw setups. Quick verdict: OpenAI is back in the coding race. live vibe check here: xitter.twitterapi.xyz/i/broadcasts/1yKAPMwWp…
1
63
7,549
GPT-5.3 lived for 24 hours. OpenAI abandoned it the moment 5.4 arrived. Ironically, they claim 5.4 has “better context retention,”but when an ongoing conversation gets hijacked, the model is forcibly replaced, and the entire context collapses, who exactly needs context retention then? They called 4o “too sycophantic.” They called 5.2 “cringe.” What will they call 5.3 now → yesterday’s mistake? Users aren’t leaving because the models are bad. They’re leaving because you’ve made it impossible to trust the model at all.
Image from 𝘊𝘰𝘳𝘳𝘪𝘯𝘦 (@OopsGuess)
4
62
5,802
Wen GPT-5.5?
3
59
2,334
I’ve been testing it in early access for a while now, and let me tell you... this model is GOOD xitter.twitterapi.xyz/flavioAd/status/202962…
Flavio Adamo (@flavioAd) profile pictureFlavio Adamo
@flavioAdMar 5
🚨 GPT-5.4 IS OUT!! I’ve been testing it in early access for a while and I’ve been using it nonstop, absolutely loving it - 1M context window - #1 on SWE-Bench — 57.7 % - 5.4 Pro #1 on FrontierMath Just to give you an idea: we had a big Avely update planned for late March because a few parts of the site were still too time-consuming to pull off with previous models GPT-5.4 basically one-shotted them on codex, so we’re moving it up to this week... this model is GOOD
Image from Flavio Adamo (@flavioAd)
3
51
4,021
Here we go
2
1
47
3,575
OpenAI has been genuinely cooking, when is GPT 5.4 Codex going to be released ?
1
35
15,798
thank you Roon ❤️
1
33
1,299
Very nice guys.
1
1
25
2,297
added to scira.ai! xitter.twitterapi.xyz/zaidmukaddam/status/20…
Zaid (@zaidmukaddam) profile pictureZaid
@zaidmukaddamMar 5
GPT 5.4, GPT 5.4 Thinking and GPT 5.4 Thinking xHigh are now available in scira.ai!
Image from Zaid (@zaidmukaddam)
25
2,123
Meanwhile GPT 5.3
Image from AVB (@neural_avb)
2
23
2,367
A new week in AI.
Image from Alex Northstar (@NorthstarBrain)
3
2
1
19
2,699
@grok what’s the hallucination rate improvement
1
19
11,837
GPT-5.4 is available on @Bolt__AI now ⚡
Image from Daniel Nguyen (@daniel_nguyenx)
3
18
2,275
@theo best model in the world counter +1
15
2,059
youtube.com/watch?v=pvO2xUb1…
2
1
9
545
@OpenAI
Image from piet (@repiet)
14
1,466
@OpenAI
Image from Bearly AI (@bearlyai)
3
11
833
We are here to help humanity, not to help destroy themselves.
Image from Evinstein 𝕏 (@Evinst3in)
1
1
12
971
Live and ready to prompt on Yupp! We’re excited to see how our community make great use of 5.4 – and how it performs on our leaderboards, of course! xitter.twitterapi.xyz/yupp_ai/status/2029623…
Yupp (@yupp_ai) profile pictureYupp@yupp_aiMar 5
📢 New Model Drop: GPT 5.4 is now live on Yupp! @OpenAI's latest flagship model boasts a 1M token context window. We've got standard and High versions. Excited to see what our community does with the latest GPT – and to find out how it rates on our user-preference leaderboards.
Image from Yupp (@yupp_ai)
10
807
finally a 1m context window from openai
10
874
@unbrowse maybe you should share your agentic browsing stats
1
6
559
GPT or Claude for Coding?
42%GPT
58%Claude
308 votes • Final results
1
6
2,470
lfg 🚀
9
2,058
@OpenAI
Image from Pallav Agarwal (@pallavmac)
9
4
193
18,893
Load more replies