Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

Today, we’re introducing [schema]: a harness reaching 99% RHAE with Opus 4.8 + Fable 5 and 95.35% with GPT-5.6 Sol on ARC-AGI-3 Public set. [schema] makes an LLM think like a physicist. 🧵

Haven Feng

5,241 subscribers

956,732 görüntüleme • 2 ay önce •via X (Twitter)

Bilim & Teknoloji Eğitim Haberler & Politika

Anya Rossi• Live Now

Private livecam show

40 Yorum

Haven Feng profil fotoğrafı

Haven Feng2 ay önce

ARC-AGI-3 gives an agent a 64×64 grid plus legal actions, no rules, stated goal, or reward. The agent must discover both what the world is and how it works like a physicist: 1. State grounding -> identify objects, relations, and goals. 2. Mechanism discovery -> infer how these states change.

Haven Feng profil fotoğrafı

Haven Feng2 ay önce

[schema] handles the state and mechanism in one editable program, a symbolic world model. It designs experiments to verify hypotheses, backtests the program against history, and plans inside its world at zero action cost.

Haven Feng profil fotoğrafı

Haven Feng2 ay önce

[schema]'s saturation of the ARC-AGI-3 public set is only a starting point. There is much more to explore! Full blog: Agent traces: Amazing team effort with @guanningzeng, @Jiani_Wang_, @wenjie_ma, @shaofeng_y27736, @ChenyangWa70207, @lustralisk95, @akanazawa, @wodenimoni, @xiuyu_l and @Zanette_ai

Haven Feng profil fotoğrafı

Haven Feng2 ay önce

@Jiani_Wang_ @wenjie_ma @shaofeng_y27736 @arcprize

Burny - Effective Curiosity profil fotoğrafı

Burny - Effective Curiosity2 ay önce

thinking like a physicist is all you need?

JFPuget 🇫🇷🇺🇦🇨🇦🇬🇱 profil fotoğrafı

JFPuget 🇫🇷🇺🇦🇨🇦🇬🇱2 ay önce

What is your take about overfitting to these 25 games? My experience on arc agi2 is that it is easy to overfit to public data. I don;'t know if it true for arc agi3 yet, but the jury is out.

Karol profil fotoğrafı

Karol2 ay önce

broooo this benchmark was planned to remain viable a little longer than that xD Well played

J Vijayavallabh 🧬/acc profil fotoğrafı

J Vijayavallabh 🧬/acc2 ay önce

Can you release harness as a skill for people to use in claude code and codex?

Dariusz Parzygnat profil fotoğrafı

Dariusz Parzygnat2 ay önce

ok, you beat the kimi k3 announcement. that is massive.

IBAN profil fotoğrafı

IBAN2 ay önce

Wow This is insane

Angela Dai profil fotoğrafı

Angela Dai2 ay önce

very cool :)

Haven Feng profil fotoğrafı

Haven Feng2 ay önce

Thanks!!!

LuisAlfonsoHernandez profil fotoğrafı

LuisAlfonsoHernandez2 ay önce

would be interesting to try with Kimi K3

alejandro profil fotoğrafı

alejandro2 ay önce

can we use it?

Xingyu Dang profil fotoğrafı

Xingyu Dang2 ay önce

How is this related to

Pud ⚛️🧪🐼 profil fotoğrafı

Pud ⚛️🧪🐼2 ay önce

The current verified score is 7.8% 🤯

Marko Kraemer profil fotoğrafı

Marko Kraemer2 ay önce

is the harness / repo going to be made open source?

Mojtaba Tabatabaie profil fotoğrafı

Mojtaba Tabatabaie2 ay önce

interesting, when is it available to test?

Aurel Prosz profil fotoğrafı

Aurel Prosz2 ay önce

Tokenwise how efficient is this? Congratz by the way, this is awesome!

Josh Cason profil fotoğrafı

Josh Cason2 ay önce

Is it possible there was some cheating somehow?

William Lamkin profil fotoğrafı

William Lamkin2 ay önce

would be cool to see a pluralistic world-model harness that externalizes a given user’s recurring lenses/perspectives as inspectable, competing, revisable cognitive instruments; recruits specialized agents to advocate, translate, falsify, experiment, and build; and uses surprise, provenance, and artifact production to turn an evolving personal worldview into a cumulative research institution.

Ridhvik Gopal (Vik) profil fotoğrafı

Ridhvik Gopal (Vik)2 ay önce

Wasn’t the whole point of ARC-AGI-3 to test the LLM and *NOT* the harness? What is the point here?

Michael Schwab profil fotoğrafı

Michael Schwab2 ay önce

Is this music generated by AI? I want to buy it

Haven Feng profil fotoğrafı

Haven Feng2 ay önce

Yeah! It’s made by coauthor @ChenyangWa70207 with @suno

arctic_ profil fotoğrafı

arctic_2 ay önce

The background beat is actually too fire I had to rewatch to read anything

max.jpg profil fotoğrafı

max.jpg2 ay önce

this is banger congrats

Jonathan Tien profil fotoğrafı

Jonathan Tien2 ay önce

This might be an over simplification, but sounds like instead of training a world model, the schema harness orients the LLM to help construct a world model on the fly?

Julian C profil fotoğrafı

Julian C2 ay önce

Wait, 99% RHAE? That’s a massive leap. What’s the secret sauce in [schema] that makes it click?

Sriraam profil fotoğrafı

Sriraam2 ay önce

Waiting for private set scores now

Thibaud profil fotoğrafı

Thibaud2 ay önce

Just incredible. Thank you so much for sharing.

kern profil fotoğrafı

kern2 ay önce

This is crazy!

Ishaan profil fotoğrafı

Ishaan2 ay önce

This is awesome

Andreas profil fotoğrafı

Andreas2 ay önce

Amazing presentation (and results).

Conor profil fotoğrafı

Conor2 ay önce

congrats!

lowkmessi profil fotoğrafı

lowkmessi2 ay önce

nice presentation

chris_paul_walker profil fotoğrafı

chris_paul_walker2 ay önce

Congrats can’t wait to check it out

Guillermo Barbadillo profil fotoğrafı

Guillermo Barbadillo2 ay önce

Very nice work, congratulations! One detail that I missed in the blog post is the cost of running the evaluation. Thanks!

Stefan Larsen profil fotoğrafı

Stefan Larsen2 ay önce

What about Fable/Sol-Max + Luna-Max? In terms of cost-efficiency...

SoniP_Gems🇮🇳 profil fotoğrafı

SoniP_Gems🇮🇳2 ay önce

Hey bro this was only planned for continue viable for long duration.

Stanislav Sorokin profil fotoğrafı

Stanislav Sorokin2 ay önce

What does 99% prove if the harness rewards behavior production does not need? The judge and operating spec must share one pass condition. Practical gate:

Benzer Videolar

GPT 5.6 Sol & Terra just made Fable 5's pricing look like a joke! We gave OpenAI's top models (Sol, Terra) and Anthropic's top models (Fable 5, Opus 4.8) the same 3 prompts: • Supernova boom • Meteor hitting a city • Solar system model The bill: Sol: $4.77 Terra: $1.24 Fable 5: $9.94 Opus 4.8: $2.46 Outputs came out surprisingly close. The prices didn't. Fable 5 cost unreasonably more than everything else, with Sol not far behind. Considering latest Fable 5 nerf, it's hard to see what you're paying for.

GPT 5.6 Sol & Terra just made Fable 5's pricing look like a joke! We gave OpenAI's top models (Sol, Terra) and Anthropic's top models (Fable 5, Opus 4.8) the same 3 prompts: • Supernova boom • Meteor hitting a city • Solar system model The bill: Sol: $4.77 Terra: $1.24 Fable 5: $9.94 Opus 4.8: $2.46 Outputs came out surprisingly close. The prices didn't. Fable 5 cost unreasonably more than everything else, with Sol not far behind. Considering latest Fable 5 nerf, it's hard to see what you're paying for.

AI/ML API

149,158 görüntüleme • 2 ay önce

Kimi K3 beating Fable and Sol on price and quality. Command Code's internal flappy bench results are in. Really impressed with the results. For Kimi K3 we set the gameplay to hard while Fable 5 and GPT-5.6 Sol were set to easier. Cost breakdown: • Kimi K3: $0.024 •Fable 5: $0.42 •GPT-5.6 Sol: $0.15 Ranking based on DX & features: •Kimi K3: 8/10 •Fable 5: 7.5/10 • GPT-5.6 Sol: 7/10 GPT 5.6 Sol max can definitely go on longer and wins the game overall, but that's something different. Grok 4.5 beats GPT 5.6 Sol on that, lol.Mm-hmm.

Kimi K3 beating Fable and Sol on price and quality. Command Code's internal flappy bench results are in. Really impressed with the results. For Kimi K3 we set the gameplay to hard while Fable 5 and GPT-5.6 Sol were set to easier. Cost breakdown: • Kimi K3: $0.024 •Fable 5: $0.42 •GPT-5.6 Sol: $0.15 Ranking based on DX & features: •Kimi K3: 8/10 •Fable 5: 7.5/10 • GPT-5.6 Sol: 7/10 GPT 5.6 Sol max can definitely go on longer and wins the game overall, but that's something different. Grok 4.5 beats GPT 5.6 Sol on that, lol.Mm-hmm.

Ahmad Awais

255,237 görüntüleme • 2 ay önce

Crazy. Opus 5 outperformed Fable 5 on 3d modeling at 2/3 the price We tested Opus 5, Fable 5, and GPT 5.6 Sol on the same prompt: build an explorable Japanese suburban street in Three.js, full 3D, rendered like a hand-drawn anime background. All on Max effort. Opus 5 took the longest but delivered the most varied scenery: clean dark outlines, flat cel-shaded color bands, dense street details down to the vending machine signage. Fable paid more attention to details, and while GPT 5.6 Sol was 3-4x cheaper and the fastest, but the render shows it: sparse, washed out, barely styled. Opus 5: 1hr 21min · $26.7 Fable 5: 1hr 5min · $35 GPT 5.6 Sol: 41min · $7.1

Crazy. Opus 5 outperformed Fable 5 on 3d modeling at 2/3 the price We tested Opus 5, Fable 5, and GPT 5.6 Sol on the same prompt: build an explorable Japanese suburban street in Three.js, full 3D, rendered like a hand-drawn anime background. All on Max effort. Opus 5 took the longest but delivered the most varied scenery: clean dark outlines, flat cel-shaded color bands, dense street details down to the vending machine signage. Fable paid more attention to details, and while GPT 5.6 Sol was 3-4x cheaper and the fastest, but the render shows it: sparse, washed out, barely styled. Opus 5: 1hr 21min · $26.7 Fable 5: 1hr 5min · $35 GPT 5.6 Sol: 41min · $7.1

GMI Cloud

54,140 görüntüleme • 1 ay önce

This is a tutorial a 12 min on how Build 3D Websites in Minutes with Al (0 Design + 0 Coding) with Claude Fable 5 and Opus 4.8 Claude Fable 5 and Opus 4.8 are massive

This is a tutorial a 12 min on how Build 3D Websites in Minutes with Al (0 Design + 0 Coding) with Claude Fable 5 and Opus 4.8 Claude Fable 5 and Opus 4.8 are massive

FHILY👑

49,063 görüntüleme • 3 ay önce

Announcing Ship: an endpoint with the highest intelligence per dollar of any frontier model. Today, Ship makes using Opus and GPT 5.6 Sol 50% cheaper guaranteed by a quality SLA.

Announcing Ship: an endpoint with the highest intelligence per dollar of any frontier model. Today, Ship makes using Opus and GPT 5.6 Sol 50% cheaper guaranteed by a quality SLA.

Martian

656,428 görüntüleme • 2 ay önce

Tested FlappyBench with GLM 5.3, Fable 5 and GPT-5.6 Sol 3 models. Same prompt with /design command. Scored on features, UX/UI, and cost. 🔹 Fable 5 → 9.5/10 · $0.420 🔹 GLM 5.3 → 9/10 · $0.018 🔹 GPT-5.6 Sol → 9/10 · $0.150 Results: → Fable 5 wins on quality, but costs 23x more than GLM 5.3 → GLM 5.3 gives better output than GPT 5.6 Sol at 8x cheaper → Optimizing for cost? Go for GLM 5.3. Otherwise, Fable 5 if quality matters

Tested FlappyBench with GLM 5.3, Fable 5 and GPT-5.6 Sol 3 models. Same prompt with /design command. Scored on features, UX/UI, and cost. 🔹 Fable 5 → 9.5/10 · $0.420 🔹 GLM 5.3 → 9/10 · $0.018 🔹 GPT-5.6 Sol → 9/10 · $0.150 Results: → Fable 5 wins on quality, but costs 23x more than GLM 5.3 → GLM 5.3 gives better output than GPT 5.6 Sol at 8x cheaper → Optimizing for cost? Go for GLM 5.3. Otherwise, Fable 5 if quality matters

Command Code

22,840 görüntüleme • 1 ay önce

We ran a test between the new Qwen3.8-Max, Opus 5 and GPT-5.6 Sol. 3 models. same prompt. one-shot with the /design command. Reviewed gameplay features, UX/UI and cost. 🔹 Qwen3.8-Max → 9/10 · $0.0248 🔹 GPT-5.6 Sol → 9/10 · $0.150 🔹 Opus 5 → 8.5/10 · $0.253 Qwen3.8-Max is approximately 4.2× cheaper than GPT-5.6, Sol, Opus 5 and has the same level of UI, UX, and gameplay.

We ran a test between the new Qwen3.8-Max, Opus 5 and GPT-5.6 Sol. 3 models. same prompt. one-shot with the /design command. Reviewed gameplay features, UX/UI and cost. 🔹 Qwen3.8-Max → 9/10 · $0.0248 🔹 GPT-5.6 Sol → 9/10 · $0.150 🔹 Opus 5 → 8.5/10 · $0.253 Qwen3.8-Max is approximately 4.2× cheaper than GPT-5.6, Sol, Opus 5 and has the same level of UI, UX, and gameplay.

Command Code

179,535 görüntüleme • 1 ay önce

Kimi K3 just 3 shotted a whole walkable voxel town with drivable cars — Voxelville, around $3.24 in API usage. The same token cost would be $10.80 with Fable 5 and $6 with GPT 5.6 Sol. Fable 5 built its own Maplewood, GPT 5.6 Sol built Willowwick, but on price/quality Kimi K3 just cooked both of them. The era of free indie game development is closer than you think!

Kimi K3 just 3 shotted a whole walkable voxel town with drivable cars — Voxelville, around $3.24 in API usage. The same token cost would be $10.80 with Fable 5 and $6 with GPT 5.6 Sol. Fable 5 built its own Maplewood, GPT 5.6 Sol built Willowwick, but on price/quality Kimi K3 just cooked both of them. The era of free indie game development is closer than you think!

Kirill

65,087 görüntüleme • 2 ay önce

Introducing autoresearch with GPT 5.6 We had GPT 5.6 Sol reproduce the key findings from “Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in LLM Finetuning” Compared to GPT-5.5 and even Fable 5, GPT-5.6 stayed more focused on a few, critical experiments and spent less time on peripheral details. It also asked fewer “clarification” questions and independently resolved ambiguities instead of pushing them back to us OpenAI pushing the boundaries of the automated research loop with models that don’t have handcuffs

Introducing autoresearch with GPT 5.6 We had GPT 5.6 Sol reproduce the key findings from “Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in LLM Finetuning” Compared to GPT-5.5 and even Fable 5, GPT-5.6 stayed more focused on a few, critical experiments and spent less time on peripheral details. It also asked fewer “clarification” questions and independently resolved ambiguities instead of pushing them back to us OpenAI pushing the boundaries of the automated research loop with models that don’t have handcuffs

alphaXiv

88,240 görüntüleme • 2 ay önce

GPT-5.6 Sol vs Fable 5 vs Grok 4.5 Built a car racing game with a UX-over-UI focus, using /design command. Parameters we reviewed: Car speed, crash behavior, scoring, smoothness, and feel. Results: Grok 4.5: best overall performance Fable 5: a close second, nearly matching Grok 4.5 GPT-5.6 Sol: great UI, weak gameplay (excess drift and friction) Ranking (based on actual gameplay): - Grok 4.5: 9/10 - Fable 5: 8/10 - GPT-5.6 Sol: 7/10 (UI was 10/10, gameplay wasn't)

GPT-5.6 Sol vs Fable 5 vs Grok 4.5 Built a car racing game with a UX-over-UI focus, using /design command. Parameters we reviewed: Car speed, crash behavior, scoring, smoothness, and feel. Results: Grok 4.5: best overall performance Fable 5: a close second, nearly matching Grok 4.5 GPT-5.6 Sol: great UI, weak gameplay (excess drift and friction) Ranking (based on actual gameplay): - Grok 4.5: 9/10 - Fable 5: 8/10 - GPT-5.6 Sol: 7/10 (UI was 10/10, gameplay wasn't)

Command Code

79,680 görüntüleme • 2 ay önce

Grok 4.5 is a "Wow!" moment. 🤯 > Opus 4.8 & GPT-5.5 level on benchmarks > Costs less than Sonnet 5 Is that real? Hard to believe. Added Grok 4.5 as an implementer agent in fable-advisor. It replaces Sonnet 5. Use Fable as the orchestrator and a combo of: 1. Grok 4.5 2. Opus 4.8 3. GPT-5.5 As implementer agents. Uses the Grok Build CLI so if you log in with an X premium account you don't have to pay API fees.

Grok 4.5 is a "Wow!" moment. 🤯 > Opus 4.8 & GPT-5.5 level on benchmarks > Costs less than Sonnet 5 Is that real? Hard to believe. Added Grok 4.5 as an implementer agent in fable-advisor. It replaces Sonnet 5. Use Fable as the orchestrator and a combo of: 1. Grok 4.5 2. Opus 4.8 3. GPT-5.5 As implementer agents. Uses the Grok Build CLI so if you log in with an X premium account you don't have to pay API fees.

Dan McAteer

191,618 görüntüleme • 2 ay önce

We ran a test between Opus 5, GPT-5.6 Sol, and Kimi K3. One prompt. Across all three models. With our /design command. Results: We reviewed gameplay, UX, and UI 🔹 Opus 5 is strong UI + motion, best gameplay 🔹 Kimi K3 has solid design and gameplay 🔹 GPT-5.6 Sol has best UI, worse gameplay Ranking (DX, features, and cost): 🔹 Opus 5: 10/10 · $0.25 🔹 Kimi K3: 9.5/10 · $0.12 🔹 GPT-5.6 Sol: 7/10 · $0.32

We ran a test between Opus 5, GPT-5.6 Sol, and Kimi K3. One prompt. Across all three models. With our /design command. Results: We reviewed gameplay, UX, and UI 🔹 Opus 5 is strong UI + motion, best gameplay 🔹 Kimi K3 has solid design and gameplay 🔹 GPT-5.6 Sol has best UI, worse gameplay Ranking (DX, features, and cost): 🔹 Opus 5: 10/10 · $0.25 🔹 Kimi K3: 9.5/10 · $0.12 🔹 GPT-5.6 Sol: 7/10 · $0.32

Command Code

24,686 görüntüleme • 1 ay önce

wow... GPT 5.6 Sol Passed the Replit Benchmark. In one prompt 5.6 on Codex created a full Replit clone with DB, Sandboxes, and an agent that builds and edits full web apps. Only Fable and GPT 5.6 have passed this benchmark. 5.6 is much better at designing iOS apps than 5.5 as well.

wow... GPT 5.6 Sol Passed the Replit Benchmark. In one prompt 5.6 on Codex created a full Replit clone with DB, Sandboxes, and an agent that builds and edits full web apps. Only Fable and GPT 5.6 have passed this benchmark. 5.6 is much better at designing iOS apps than 5.5 as well.

Riley Brown

123,242 görüntüleme • 2 ay önce

Gave a random doodle to Kimi K3, Fable 5, Opus 5, GPT-5.6 Sol and asked them to finish the drawing 🎨 The only rule: make it as creative as possible. Here are the results 👇 Which one do you like most?

Gave a random doodle to Kimi K3, Fable 5, Opus 5, GPT-5.6 Sol and asked them to finish the drawing 🎨 The only rule: make it as creative as possible. Here are the results 👇 Which one do you like most?

TypingMind

289,842 görüntüleme • 1 ay önce

Found a way to get $125-$200 in FREE ai credits for gpt-5.6 sol, claude opus 5 and opus 4.8 😳 agentrouter is a non-profit version of openrouter. one api key, no card. credits land on your account at signup. what you get for $0: -$125-200 in credits on signup (github only) -claude opus 5, gpt-5.6 sol, opus 4.8 through one key -free models -openai-compatible (works with anything) what this replaces: -paid frontier API costs for coding agents -claude opus 4.8: $15/M in, $75/M out -gpt-5.6 sol: $25/M out how to grab yours (2 min): 1. go to → sign in with github (important) 2. console → create your api key 3. base url: 4. pick a model and go Works in hermes, claude code, cursor, aider and any openai-compatible tools Important: github-only signup, older accounts get the full credit, fresh ones may get less. credits are one-time. referral adds +$100 use this before they tighten the offer

Found a way to get $125-$200 in FREE ai credits for gpt-5.6 sol, claude opus 5 and opus 4.8 😳 agentrouter is a non-profit version of openrouter. one api key, no card. credits land on your account at signup. what you get for $0: -$125-200 in credits on signup (github only) -claude opus 5, gpt-5.6 sol, opus 4.8 through one key -free models -openai-compatible (works with anything) what this replaces: -paid frontier API costs for coding agents -claude opus 4.8: $15/M in, $75/M out -gpt-5.6 sol: $25/M out how to grab yours (2 min): 1. go to → sign in with github (important) 2. console → create your api key 3. base url: 4. pick a model and go Works in hermes, claude code, cursor, aider and any openai-compatible tools Important: github-only signup, older accounts get the full credit, fresh ones may get less. credits are one-time. referral adds +$100 use this before they tighten the offer

painn

109,551 görüntüleme • 1 ay önce

GPT-5.6 Sol has been used to solve open problems in mathematics. So why was it struggling with ARC-AGI-3, a benchmark of 2D puzzle games? We investigated. The harness was not letting it remember what it had learned. We found that enabling two API settings tripled our scores with 6x fewer output tokens.

GPT-5.6 Sol has been used to solve open problems in mathematics. So why was it struggling with ARC-AGI-3, a benchmark of 2D puzzle games? We investigated. The harness was not letting it remember what it had learned. We found that enabling two API settings tripled our scores with 6x fewer output tokens.

OpenAI

1,239,835 görüntüleme • 1 ay önce

I gave Kimi K3, Gemini 3.7 Flash, GPT 5.6 Sol, and Claude Opus 5 a dice and asked them to draw a herbivorous animal integrated with it here's how they imagined the animal fused with the dice

I gave Kimi K3, Gemini 3.7 Flash, GPT 5.6 Sol, and Claude Opus 5 a dice and asked them to draw a herbivorous animal integrated with it here's how they imagined the animal fused with the dice

Ann Nguyen

411,939 görüntüleme • 28 gün önce

🚨 Create Your Own Custom Coding Agent - Mix Fable 5, GPT 5.6 sol and Grok 4.5! Our newest feature allows you to mix and match your favorite LLMs and create your custom coding agent Create your favorite cocktail such as - Fable 5 for hard coding - 5.6 sol for backend - Opus 4.8 for frontend - Grok 4.5 for easy coding You can also only choose cheap LLMs for a cost optimized experience. Your use custom agent in API, chat or our agent platform

🚨 Create Your Own Custom Coding Agent - Mix Fable 5, GPT 5.6 sol and Grok 4.5! Our newest feature allows you to mix and match your favorite LLMs and create your custom coding agent Create your favorite cocktail such as - Fable 5 for hard coding - 5.6 sol for backend - Opus 4.8 for frontend - Grok 4.5 for easy coding You can also only choose cheap LLMs for a cost optimized experience. Your use custom agent in API, chat or our agent platform

Bindu Reddy

7,808,123 görüntüleme • 2 ay önce

this is f**king insane Fable 5 goes away in 24 hours from Claude Code and someone just figured out how to get fable 5 reasoning in GPT 5.6 Sol with one prompt you have less than 24 hours to set this up as this takes away 20% of your usage limit here is how: 1) Paste the prompt and ask it to "make an operating manual" 2) Go the Codex App > Settings > Personalisation > Custom Instructions 3) Use GPT 5.5 / GPT 5.6 and now you have Fable 5 reasoning at 1/3 th the price save and bookmark it no matter what

this is f**king insane Fable 5 goes away in 24 hours from Claude Code and someone just figured out how to get fable 5 reasoning in GPT 5.6 Sol with one prompt you have less than 24 hours to set this up as this takes away 20% of your usage limit here is how: 1) Paste the prompt and ask it to "make an operating manual" 2) Go the Codex App > Settings > Personalisation > Custom Instructions 3) Use GPT 5.5 / GPT 5.6 and now you have Fable 5 reasoning at 1/3 th the price save and bookmark it no matter what

Avid

57,544 görüntüleme • 2 ay önce

GPT-5.6 Sol Ultra lost to GPT-5.5 on physics at 3× the cost! We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos Prompts: - A monster truck backflipping onto a parked car - A stunt car jumping six buses into a brick wall - A train derailing off a broken bridge into the water Outputs: GPT-5.6 Sol Ultra: 32.9K tokens, $0.33 Opus 4.8: 9.2K tokens, $0.24 GPT-5.5: 12.4K tokens, $0.11 Grok 4.5: 7.0K tokens, $0.08 Sol Ultra draws just like GPT-5.5, only with more detail and it shows clearest on the bus jump where the two look almost identical. In the other tests Sol Ultra came out worse than its predecessor and the physics really got weaker. We think GPT-5.5 took the truck flip and the train outright. With Sol Ultra you basically get GPT-5.5 with weaker physics and a nicer picture for 3x the price. Newborn Grok 4.5 failed two of the three tests and only came good on the train

GPT-5.6 Sol Ultra lost to GPT-5.5 on physics at 3× the cost! We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos Prompts: - A monster truck backflipping onto a parked car - A stunt car jumping six buses into a brick wall - A train derailing off a broken bridge into the water Outputs: GPT-5.6 Sol Ultra: 32.9K tokens, $0.33 Opus 4.8: 9.2K tokens, $0.24 GPT-5.5: 12.4K tokens, $0.11 Grok 4.5: 7.0K tokens, $0.08 Sol Ultra draws just like GPT-5.5, only with more detail and it shows clearest on the bus jump where the two look almost identical. In the other tests Sol Ultra came out worse than its predecessor and the physics really got weaker. We think GPT-5.5 took the truck flip and the train outright. With Sol Ultra you basically get GPT-5.5 with weaker physics and a nicer picture for 3x the price. Newborn Grok 4.5 failed two of the three tests and only came good on the train

atomic.chat

898,646 görüntüleme • 2 ay önce