What the internet is saying about GPT-4o

What the internet is saying about "GPT-4o OpenAI"

Most of the evidence is about GPT-4o's retirement and the migration to the GPT-6 family. Direct opinion on GPT-4o itself is thin. Several snippets are truncated, so I note below where details are cut off.

🟢 Supporters

• •Still the "flagship" in a gateway. An open-connect PR maps its high alias ("flagship quality") to gpt-4o, while deep maps to o3-mini [2]. This is a design choice, not praise, but it shows GPT-4o still being treated as a top-tier option.
• •Used as a benchmark reader. The agentos docs record a matched gpt-4o reader run at 84.2% on LongMemEval-S (run 18-41) [8]. In the router setup, gpt-4o answers the temporal-reasoning and single-session-user questions [6]. The 85.6% headline is a reader-router result, not a pure gpt-4o result [6].
• •No enthusiastic endorsements. I found no first-person praise of GPT-4o in the evidence.

🔴 Skeptics

• •Sunset pressure. OpenAI announced the retirement on October 14, 2026, and one team labelled the migration "P1" [1]. GPT-4o mini leaves the API on January 14, 2027 [1][3].
• •Weak quality for GPT-4o mini. In Silo's translation test, gpt-4o-mini was "the weakest OpenAI model tried" and returned 7 of 24 subtitle batches with cues missing [4]. Silo's PR defaults the text model to gpt-6-luna instead [4].
• •Reliability and quota problems. A production classroom feature paused because GPT-4o mini returned a no-credits HTTP 429, alongside Gemini rate limits [7]. The fix registers other Gemini models as fallbacks [7].
• •Migration friction. gpt-6-luna rejects max_tokens and temperature: 0.7, which the current gpt-4o-mini call sends. Callers need max_completion_tokens and no temperature [3].
• •Examples being scrubbed. NVIDIA NeMo Guardrails has a PR replacing retired OpenAI models with gpt-6-luna, though it is deferred to a future release [11].

⚪ Neutral Observers

• •Pricing tracked. Langfuse documented an "Oct 23 2026 OpenAI retirement batch" in its pricing data [5]. The snippet doesn't say which models are in it.
• •Official guidance. The Hacker News post links OpenAI's help article "Migrating from GPT-4x to GPT-6 Luna" [12], which names Luna as the successor path.
• •GPT-6 naming. Hacker News discussion centers on the newer lineup: Luna (small), Terra (middle) and Sol (large) [18]. Commenters call Terra "a weird middle child" and are confused by the naming [18]. GPT-4o is not discussed there.
• •Unrelated "GPT" items. Several items (GPT-6 Astra safety and testing, Rosalind, Cyber, political-bias and gender-bias pieces) don't concern GPT-4o [17][19][20][24][26][27]. GPT-4 Turbo's lack of Structured Outputs support also comes up, but that is a different model [9].

Cross-Platform Signals

• •GitHub dominates. Almost all GPT-4o-specific evidence is from GitHub PRs and issues, which are practical migration work [1][3][4][5][7][11].
• •Hacker News is quiet on GPT-4o. The retirement post has 4 points and 1 comment [12]. The GPT-6.1 Terra thread has 638 points and 428 comments [18].
• •Consensus. Both platforms treat GPT-4o as legacy, with attention moving to the GPT-6 line. There is no visible defense of keeping it beyond retirement.
• •Divergence. GitHub shows a small number of projects still depending on gpt-4o for benchmarks or aliases [2][6][8].

Community Verdict

The internet is treating GPT-4o, and especially GPT-4o mini, as a model on its way out. Developers are migrating to gpt-6-luna ahead of the January 14, 2027 retirement, and some report quality, quota and parameter-compatibility problems with the old model. A few projects still use gpt-4o as a reference or "flagship" reader, but the evidence has no strong advocacy for it. Attention has moved to the GPT-6 family, and the evidence for this brief is small and skewed toward GitHub.

---

Based on GitHub + Hacker News data from the last 30 days, researched with last30days by Matt Van Horn.