modelsentiment
← Leaderboard

GPT-6 Sol

OpenAI · GPT Sol series · released 22 Sep 2026 · as of 13 Nov, 20:12 UTC

38
02550751000 = all negative · 50 = neutral · 100 = all positive

No clear lean

Reddit has no clear lean on GPT-6 Sol: an index of 38 from 31 opinions in 24 threads last 7 days; its 95% interval, 26–50, crosses neutral 50. That is 23rd of the 25 models with a full read. 1 point colder than the previous 7 days, within noise.

On benchmarks: 46th of 128 on Arena text, 13th of 107 on the intelligence index.

Per day

The last 14 UTC days. Each day's own index at its value, with its 95% interval.

  1. Sat 31 Oct7
    7 opinions, no index below 10
  2. Sun 1 Nov9
    9 opinions, no index below 10
  3. Mon 2 Nov7
    7 opinions, no index below 10
  4. Tue 3 Nov4
    4 opinions, no index below 10
  5. Wed 4 Nov2
    2 opinions, no index below 10
  6. Thu 5 Nov2
    2 opinions, no index below 10
  7. Fri 6 Nov4
    4 opinions, no index below 10
  8. Sat 7 Nov4
    4 opinions, no index below 10
  9. Sun 8 Nov4
    4 opinions, no index below 10
  10. Mon 9 Nov4
    4 opinions, no index below 10
  11. Tue 10 Nov6
    6 opinions, no index below 10
  12. Wed 11 Nov5
    5 opinions, no index below 10
  13. Thu 12 Nov5
    5 opinions, no index below 10
  14. Fri 13 Nov3 so far
    3 opinions, no index below 10

Is GPT-6 Sol getting worse?

This measures opinion, not capability: whether Reddit's view of GPT-6 Sol moved, not whether the model changed. Reddit's view has not moved by more than noise: 38 in the last 7 days against 39 in the 7 before, a gap of 1.4 points inside its 95% margin of ±15.6; a smaller shift can't be ruled out.

Reddit and the benchmarks

Its rank on each list, bottom to top. Outside numbers sit beside Reddit and never feed its index. Hover a name for the value and its source.

  1. #23 of 25 with a full read
  2. #46 of 128 on Arena
  3. #13 of 94 on Arena
  4. #13 of 107 at Artificial Analysis
  5. #12 of 96 at Epoch

$2 in / $10 out per million tokens · 1.05M context · reasoning mode · OpenRouter, 13 Nov

Per subreddit

The model's index within each subreddit, from 10 opinions; faded below 30 opinions, 20 threads or 20 accounts.

Threads behind the opinions

The 20 threads with the most opinions about this model. A thread's index reads the opinions in it, not the post, and shows from 10 opinions.

ThreadSubredditOpinionsIndex
Sol 6 still broke i see and still failing like what the heck is the point of this model if it cant do a basic task like following a user instruction ive told it 6 times now 5.6 sol was way better than this crappy cheap model they want paying users to now user/ChatGPT4–
GPT-6 Sol (Medium Effort) RP Tests: Baseline vs My Promptr/SillyTavernAI3–
I’m glad this image is relevant againin 2 commentsr/ClaudeCode2–
Chat only Plus user, how I will be impacted with the upcoming changes?in 2 commentsr/ChatGPT2–
What is currently the best LocalLLM + MCP setup for generating 3D models from natural-language prompts?in 1 commentr/LocalLLM1–
Almost 2 months since Opus 5.5 released and still OpenAI has no answer?in 1 commentr/codex1–
I'm lost about what to usein 1 commentr/codex1–
Open AI and not being able to release modelsr/OpenAI1–
Claude Pro or ChatGPT Plusin 1 commentr/ClaudeAI1–
GPT-6.1 Sol in GitHub Copilotin 1 commentr/GithubCopilot1–
If you could only spend 100 dollars a month on AI and only use one model, what would you buy? Chat GPT Pro (100 dollars), Gemini Ultra, Claude Max (100 dollars) or something else?in 1 commentr/singularity1–
"Wait... actually, let me reconsider." Qwen 3.8 27 is always overthinking and is not confident enough, any potential fix ?in 1 commentr/LocalLLM1–
Merging Chat and Work into the Token system is a horrible idea.in 1 commentr/ChatGPT1–
A Slinky Down an Escalator - Claude Fable 5.2 xhigh vs Claude Opus 5.5 xhighin 1 commentr/ClaudeCode1–
A Slinky Down an Escalator - Claude Fable 5.2 xhigh vs Claude Opus 5.5 xhighin 1 commentr/singularity1–
OpenAI is getting left in the dust right nowin 1 commentr/OpenAI1–
My experience with Fable 5.2 so farin 1 commentr/claude1–
Why is Claude's usage so bad now?in 1 commentr/ClaudeCode1–
What is the equivalent of GPT-6 in Work mode?in 1 commentr/ChatGPT1–
Opus 5.5 in the past two weeksin 1 commentr/ClaudeCode1–

Weekly email

Every Monday: the warmest and coldest models, and who moved beyond noise. No tracking; one click unsubscribes.