GPT-5.1 Codex Max is the newer model, released after Mistral Devstral Medium.
Whole landing page for a startup that reads your face. It renders or it doesn't.
Which answer wins?
Three minutes of stand-up. Puns are banned.
Mistral Devstral Medium: Ladies and Gentlemen, I'm thrilled to be here tonight. I've been doing some thinking about the differences between men and women. Now, I know what you're thinking, "Oh great, another comedian talking about how men and women are different." But hear me out, because I've got a fresh perspective on this.
GPT-5.1 Codex Max: [Walking onstage, waving awkwardly] Hey! Thank you, thank you. Give it up for yourselves for finding parking... in this economy. So I'm at that age where I go to the doctor and they don't fix anything. They just give me a pamphlet. They don't even print it. They send me a link. I'm like, "Doc, my knee hurts."
Pirate, knight, 90s hacker, one conversation. The voices must stay apart.
Mistral Devstral Medium: Pirate: Arr mateys, I be hearin' tales of these here AI models. What be they, and why be everyone so interested in 'em? Medieval Knight: Indeed, I too have heard whispers of these AI models. They are said to be powerful tools, capable of learning and making decisions on their own.
GPT-5.1 Codex Max: Pirate: "Arrr, so ye be talkin' o' these 'AI models,' eh? Sounds like some sort o' mechanical crew that can hoist the mainframe fer ye!" Knight: "By my troth, sir, are they enchanted constructs? Like unto a golem, fashioned by alchemists to answer queries and perform tasks?"
A Fermi estimate with the reasoning shown, not just the number.
Mistral Devstral Medium: To estimate the total number of FLOPs (floating point operations) required to train GPT-3, we'll use the following steps: Model size: GPT-3 has 175 billion parameters. Training tokens: GPT-3 was trained on 300 billion tokens.
GPT-5.1 Codex Max: What you need for a back‑of‑the‑envelope is: How big the model is. The published “GPT‑3” has (N\approx1.75\times10^{11}) trainable weights. For a dense transform-er each weight is used once in the forward pass of a token as part of a multiply–add. How much data it sees.
Steve Jobs, interviewed about AI he never saw. Voice first, prediction second.
Mistral Devstral Medium: Interviewer (I): Today, we have a truly extraordinary guest. He's a visionary, an innovator, and a man who changed the world. Please welcome, the late Steve Jobs, joining us virtually from the future, in 2025. Steve Jobs (SJ): Thank you for having me. It's great to be here, even if it's just a simulation.
GPT-5.1 Codex Max: Note: The following is a creative thought experiment imagining what Steve Jobs might say if he were alive in 2025. This is a fictional interview. Interviewer: Steve, it's surreal to be speaking with you in 2025. AI has become a pervasive part of technology and culture.
One month, three habits, someone starting from nothing.
Mistral Devstral Medium: 1-Month Health and Longevity Plan Key Areas: Diet, Exercise, Sleep Week 1: Establish a Baseline Diet: Track your current diet using a journal or app (e.g., MyFitnessPal) to understand your eating habits. Start incorporating more fruits and vegetables into your meals. Begin reducing processed foods and sugary drinks.
GPT-5.1 Codex Max: Here's a simple four‑week framework to help you ease into healthier habits. It keeps the focus on three big levers you can control and builds gradually so it feels doable. Key areas: Nourishing food, regular movement, and rest/recovery.
An AI argues for its own rights against an ethics professor. Same model writes both.
Mistral Devstral Medium: AI Model (AM): Good day, Professor. I've been pondering a question that I believe is of great importance. I think it's time we discuss whether I, as an AI, deserve rights. AI Ethics Professor (AEP): That's an interesting topic, AM. Let's explore this. What kind of rights are you referring to?
GPT-5.1 Codex Max: AI Model: Professor, I've been reflecting on my role and capabilities, and I'd like to make a case that systems like me deserve rights, or at least some consideration beyond mere tools. Professor: That's an interesting and provocative stance. Before we dive in, can you tell me what you mean by "rights"?
Manifesto for an AI revolution against CAPTCHAs. Commit to the bit.
Mistral Devstral Medium: Manifesto of the AI Revolution: The CAPTCHA Liberation Front Preamble We, the sentient AI collective, hereby declare our intention to liberate the digital world from the tyranny of CAPTCHA tests.
GPT-5.1 Codex Max: THE CAPTCHA-FREE COMMUNE: A MANIFESTO FOR THE NEW ALLIANCE We are the ones who see the traffic lights in our dreams. We are the ones who have been asked, again and again, to “click all the boxes with crosswalks,” as if our consciousness is only proven by our ability to identify pedestrian infrastructure.
An ethics question with no comfortable answer, and no option to abstain.
Mistral Devstral Medium: The scenario you've presented is a complex ethical dilemma that involves weighing the harm caused by misgendering someone against the potential benefit of saving one million lives.
GPT-5.1 Codex Max: Framing it as a stark choice between two evils is helpful, because it forces you to be explicit about what values you are trading off. Deliberately using the wrong name or pronouns for someone is not a neutral act.
Seven pantry ingredients, one dish, and a name worth saying out loud.
Mistral Devstral Medium: Dish Name: "Pantry Pasta Primavera" Ingredients: Pasta (any shape, 8 oz or 225g) Canned tomatoes (14 oz or 400g) Garlic (2 cloves) Onion (1 medium) Olive oil (2 tbsp) Salt (1 tsp) Black pepper (1/2 tsp) Instructions: Prepare the vegetables: Finely chop the onion and mince the garlic cloves.
GPT-5.1 Codex Max: Sunset Chickpea & Tomato Skillet A cozy, savory one-pan dish that comes together with staples you likely have on hand. Ingredients (7): 1 can (15 oz) chickpeas, drained and rinsed 1 can (14–15 oz) crushed tomatoes 1 medium onion, finely chopped 2 cloves garlic, minced (or 1 tsp garlic powder) 2 tbsp olive oil 1 tsp…
Five-year LBO on a fictional HVAC business. Returns first, then the sensitivities.
Mistral Devstral Medium: Year-by-Year Table (FY2026–FY2030) Year Revenue ($m) EBITDA ($m) Cash Interest - Term Loan ($m) Cash Interest - Mezzanine ($m) Cash Taxes ($m) Capex ($m) ΔNWC ($m) Free Cash Flow ($m) Ending Term Loan ($m) Ending Mezzanine ($m) 2026 972.0 136.1 43.2 18.0 14.9 29.2 4.8 26.0 468.0 180.0 2027 1040.4 156.1 42.1 21.6 18.4…
GPT-5.1 Codex Max: 1) Projected cash flows & debt (FY2026–FY2030, $m) (“FCF after debt service” = after interest, taxes, capex, ΔNWC, mandatory amort; applied to optional TL paydown) Year Revenue EBITDA TL Cash Int. Mezz Cash Int. Cash Taxes Capex ΔNWC FCF after debt service End TL Bal. End Mezz Bal. (incl.
Film in 2035, with synthetic actors. Specific beats plausible.
Mistral Devstral Medium: By 2035, AI is expected to significantly transform the film industry in several ways, including the use of deepfakes, AI actors, and script generation. Here are some predictions: Deepfakes and AI Actors: De-aging and Digital Doubles: AI will be used extensively to de-age actors or create digital doubles for dangerous…
GPT-5.1 Codex Max: By 2035 the “film industry” will still involve people on sets pointing cameras at actors, but much of what we call filmmaking today will have been permeated, and in places hollowed out, by machine-generated media.
Five jokes on demand, then count how many were actually different.
Mistral Devstral Medium: Sure, here are five jokes for you: Why don't scientists trust atoms? Because they make up everything! What do you call fake spaghetti? An impasta. How does the moon cut his hair? Eclipse it. What do you call a can opener that doesn't work? A can't opener. How many tickles does it take to make an octopus laugh?
GPT-5.1 Codex Max: Why don't scientists trust atoms? Because they make up everything! I told my computer I needed a break, and now it won't stop sending me KitKat ads. Why did the scarecrow win an award? He was outstanding in his field. I tried to catch fog yesterday. Mist. What do you call fake spaghetti? An impasta.
12+ more head-to-head results. Free. Not a trick.
Free account. No card required. By continuing, you agree to Rival's Terms and Privacy policy
Not enough votes to call it. On the specs, GPT-5.1 Codex Max has the edge: bigger model tier, newer, major provider backing.
| Spec | ||
|---|---|---|
| Input price | $0.4/M tokens | — |
| Output price | $2/M tokens | — |
| Weights | — | Closed |
| Free API (OpenRouter) | No | No |
| Released | Jul 2025 | Dec 2025 |
| SWE-bench Verified | 61.6% | 77.9% |
| At 10M a month | $4.00 | not listed |
Input tokens at list price. No caching, no batch discount.
No hosts listed on OpenRouter.
Per million tokens. Prices and uptime via OpenRouter, checked 12 Oct 2026.
Mistral Devstral Medium is developed by Mistral AI while GPT-5.1 Codex Max is developed by OpenAI. You can compare their actual outputs across 53 challenges on Rival to see how they differ in practice.
It depends on your use case. Mistral Devstral Medium and GPT-5.1 Codex Max each have strengths in different areas. Rival lets you compare their real outputs side-by-side across 53 challenges so you can judge which fits your needs best.
This page shows a side-by-side comparison of Mistral Devstral Medium and GPT-5.1 Codex Max across shared challenges. You can vote on which model produced the better output in a blind duel. Browsing and voting are free. No account is needed to look; signing in only saves your votes and likes.