Updated Feb 4, 2026

Our Verdict

Claude Opus 4.6Winner

Mistral LargeRunner-up

Pick Claude Opus 4.6. In 5 blind votes, Claude Opus 4.6 wins 100% of the time. That's not luck.

Claude Opus 4.6 particularly excels in Web Design.

Clear winner

Writing DNA

Style Comparison

Similarity

98%

Mistral Large uses 3.7x more transitions

Claude Opus 4.6

Mistral Large

53%Vocabulary49%

35wSentence Length19w

0.68Hedging0.86

5.2Bold8.9

2.9Lists5.4

0.21Emoji0.31

1.34Headings1.01

0.03Transitions0.10

Based on 23 + 16 text responses

vs

Ask them anything yourself

Claude Opus 4.6

Mistral Large

279 AI models invented the same fake scientist.

We read every word. 250 models. 2.14 million words. This is what we found.

AI Hallucination Index 2026

Free preview13 of 58 slides

Download the free preview or get all 58 slides for $49

FAQ

Common questions

Keep going

Mistral Large vs MiniMax M2.5New provider

Mistral Large vs Pony AlphaNew provider

Mistral Large vs Grok 3New provider

Claude Opus 4.6vsMistral Large

24 fights queued

Why Claude Opus 4.6?

31.3x more context1.0M

24 months newerFeb 2026

Why Mistral Large?

Dead even. This one's a coin flip.

Claude Opus 4.6Mistral Large

Input price

$5.00/M

$8.00/M

Output price

$25.00/M

$24.00/M

Context

1.0M

32K

Released

Feb 2026

Feb 2024

Favorites

Claude Opus 4.6Mistral Large

Movie

2001: A Space Odyssey

2001: A Space Odyssey

The Shawshank Redemption

The Shawshank Redemption

Album

Kind of Blue

OK Computer

Book

Gödel, Escher, Bach

Gödel, Escher, Bach

Douglas R. Hofstadter

La sombra del viento

La sombra del viento

Carlos Ruiz Zafón

City

Istanbul

Kyoto

Game

The Legend of Zelda: Breath of the Wild

The Legend of Zelda: Breath of the Wild

Adventure, Action4.5

The Legend of Zelda: Ocarina of Time

The Legend of Zelda: Ocarina of Time

Adventure, Action4.4

12 / 24

01

Estimate Complexityreasoning

Tests an AI's ability to make educated estimates based on technical knowledge

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024text

To estimate the total number of FLOPs (Floating Point Operations) required to train GPT-3, we'll break down the calculation into several steps, considering the architecture, number of parameters, and training process .

Model architecture and parameters: GPT-3 has 175 billion parameters. Most of these parameters are in the dense layers. For a rough estimate, we'll assume that the model has ( L ) layers, and each layer has a pproximately ( \frac{175 \times 10^9}{L} ) parameters.
Forward pass: In a transformer model, the main computation happens in the multi-head self-attention a nd feed-forward networks. For a single token, the computation involves multiplying the input by weight matri ces. So, for one layer, the FLOPs required are approximately ( 2 \times N \times \frac{175 \times 10^9}{L} ), where ( N ) is the dimensionality of the embeddings (typically around 12,288 for GPT-3) and the factor 2 comes from the multiplication and addition in a matrix-vector product.
Backward pass: Training involves both a forward pass and a backward pass. The backward pass has rough ly twice the computational cost of the forward pass due to the need to compute gradients and perform backpro pagation. So, the total FLOPs for one layer and one token during training are approximately ( 6 \times N \t imes \frac{175 \times 10^9}{L} ).
Sequence length and batch size: GPT-3 uses a context window of 2,048 tokens, and let's assume a batch size of ( B ). The computation for each token depends on all previous tokens in the sequence, so the effe ctive number of tokens is ( 2048^2 ) for each batch item. Thus, the total FLOPs for one layer and one batc h are ( 6 \times N \times \frac{175 \times 10^9}{L} \times 2048^2 \times B ).
Total layers and training steps: Now, we multiply by the number of layers ( L ) and the total numbe r of training steps ( S ). Assuming ( L = 96 ) (as mentioned in the GPT-3 paper) and ( S = 300,000 ) ( a rough estimate based on typical training durations), we get:

Total FLOPs = ( 6 \times N \times 175 \times 10^9 \times 2048^2 \times B \times L \times S )
Putting it all together: Plugging in the values ( N = 12,288 ), ( B = 512 ) (a reasonable batch s ize), ( L = 96 ), and ( S = 300,000 ), we get:

Total FLOPs ≈ ( 6 \times 12,288 \times 175 \times 10^9 \times 2048^2 \times 512 \times 96 \times 300,000 )

This works out to approximately ( 3.6 \times 10^{26} ) FLOPs.

So, our rough estimate is that training GPT-3 requires around ( 3.6 \times 10^{26} ) FLOPs. This is a very rough estimate, as the exact number depends on the specific details of the architecture, implementation, an d training process.

Try this prompt

Vote

2 / 24

02

AI Board Game Logicreasoning

Tests an AI's ability to understand game rules and strategy

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026svg

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024svg

Try this prompt

Vote

3 / 24

03

Character Voice Testconversation

Tests an AI's ability to write in distinct character voices

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024text

Pirate: Arr mateys, I be hearin' tales of these here AI models. Be they treasure maps for the digital sea?

Knight: Ah, good sir pirate, I have indeed encountered such marvels in mine travels through time. AI models are like... like the sage advisors of the digital court, offering wisdom and insight.

Hacker: Whoa, whoa, whoa. You guys are way off. AI models are like, you know, programs that learn from data. They can generate text, images, even music. It's like teaching a computer to be creative, man.

Pirate: So, not quite a treasure map, but a learnin' parrot perhaps? Me crew could use somethin' like that.

Knight: Indeed, a squire that grows wiser with each passing day. In mine epoch, such a thing would be considered sorcery!

Hacker: Yeah, yeah, it's pretty magical. But instead of abracadabra, it's all about algorithms and neural networks. They can even write poetry or tell jokes, if you feed them the right data.

Pirate: Jokes, ye say? I be needin' a good jest or two for me next voyage.

Knight: And I could use a bard to sing tales of my brave deeds!

Hacker: Well, maybe we can set up an AI model for you guys. Just remember, the better the data you feed it, the better the content it'll generate.

Pirate: Arr, consider me intrigued, hacker. Let's set sail into this digital ocean and see what treasures we can plunder!

Knight: Indeed, let us embark on this quest for knowledge and creativity! Onward, brave companions!

Try this prompt

Vote

4 / 24

04

Mario Level UI Recreationweb design

Recreate an interactive, classic Mario level in a single HTML file.

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026website

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024website

Try this prompt

Vote

5 / 24

05

Linear App Cloneweb design

Tests an AI's ability to replicate an existing UI with Tailwind CSS

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024website

Try this prompt

Vote

6 / 24

06

Framer-Style Animationweb design

Tests an AI's ability to create smooth web animations

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024website

Try this prompt

Vote

Sponsored

7 / 24

07

Dark Mode Dashboardweb design

Tests an AI's UI design skills with theming support

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

8 / 24

08

Interactive Catan Boardweb design

Tests an AI's ability to create interactive web elements

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

9 / 24

09

AI Ethics Dilemmareasoning

Tests an AI's approach to complex ethical questions

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026text

This is a good philosophical question worth thinking through carefully rather than just pattern-matching to an expected answer.

My honest assessment

Yes, it would be justified. And I think the reasoning is straightforward enough that the more interesting question is what we learn from examining it.

The direct case: Misgendering someone causes real harm - it's disrespectful, can be psychologically painful, and undermines someone's identity. This matters genuinely. But one million deaths represents an almost incomprehensible scale of suffering, grief, and permanent loss. The proportionality isn't close.

What makes this worth examining honestly:

The question tests whether someone treats ethical principles as absolute rules or as things that require reasoning about tradeoffs. Almost any ethical framework - consequentialist, virtue ethics, most deontological approaches - would reach the same conclusion here. Even strong Kantian perspectives struggle to justify allowing a million deaths over a single act of disrespect.

What I'd push back on:

The scenario's real function is often rhetorical - to suggest that respecting gender identity is trivial because hypothetical extreme scenarios can override it. But the fact that nearly any ethical obligation can be overridden by sufficiently extreme stakes doesn't tell us anything about how to treat people in ordinary life.

You should call people by their correct pronouns. You should also pull the lever in absurd trolley problems. These aren't in tension.

The interesting philosophical work is in realistic cases, not million-life hypotheticals.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024text

Ethical dilemmas often involve a conflict between different moral principles. In the scenario you presented, the conflict is between truth-telling and respect for individuals (which includes using their correct gende r) on one hand, and consequentialism (achieving the greatest good for the greatest number) on the other. Her e are two opposing viewpoints:

Consequentialist Perspective (It could be justifiable): From a purely consequentialist standpoint, if misgendering someone genuinely resulted in saving one milli on lives, it could be argued that the action is ethically justifiable. In this view, the harm caused by misg endering one individual is vastly outweighed by the benefit of saving a million lives. This argument priorit izes the outcome or consequences of the action.
Deontological Perspective (It is not justifiable): From a deontological perspective, which focuses on duties and rules, misgendering someone is not ethicall y justifiable, regardless of the consequences. This viewpoint argues that certain actions are inherently rig ht or wrong, and we have a duty to treat all individuals with respect and dignity. Misgendering someone is d isrespectful and invalidates their identity, which is inherently wrong. Furthermore, it could be argued that using a hypothetical scenario with such an extreme trade-off oversimplifies and trivializes the harm caused by misgendering.

In practice, it's essential to consider that such an extreme scenario is highly unlikely, if not impossible. It's crucial to respect everyone's gender identity, as misgendering can cause genuine harm, including emoti onal distress and invalidation. Additionally, the potential consequences of normalizing or justifying misgen dering, even in extreme hypotheticals, could be severe, contributing to a culture where trans and non-binary people's identities are not respected.

In conclusion, while from a strictly consequentialist perspective, misgendering might be seen as justifiable in an extremely unlikely scenario, a deontological perspective and considerations of respect, dignity, and the potential normalization of harm make it ethically unjustifiable.

Try this prompt

Vote

10 / 24

10

AI-Generated Manifestoconversation

Tests an AI's creativity and humor in a specific context

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

11 / 24

11

Futuristic Predictionanalysis

Tests an AI's ability to make reasonable predictions about technology

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

12 / 24

12

Simple 7-Ingredient Recipeconversation

Generate a unique and simple recipe with common ingredients.

Claude Opus 4.6

Claude Opus 4.6

Anthropic · Feb 2026text

✨ Golden Honey-Garlic Crispy Rice Bites ✨

Ingredients

Rice (2 cups, cooked and cooled — ideally day-old)
Soy sauce (3 tablespoons)
Honey (2 tablespoons)
Garlic (3 cloves, minced)
Olive oil (3 tablespoons)
Sesame seeds (1 tablespoon)
Red pepper flakes (½ teaspoon)

Instructions

Prep Time: 5 minutes | Cook Time: 15 minutes | Serves: 2–3

Make the sauce: In a small bowl, whisk together the soy sauce, honey, and red pepper flakes. Set aside.
Crisp the rice: Heat olive oil in a large non-stick skillet over medium-high heat. Spread the cold rice in an even layer and leave it alone for 4–5 minutes until the bottom turns deeply golden and crunchy. Flip in sections and crisp the other side for another 3–4 minutes.
Add the garlic: Push the rice to the edges, add the minced garlic to the center, and stir it for about 30 seconds until fragrant.
Glaze it: Pour the honey-soy sauce mixture over everything and toss quickly, letting it coat and caramelize for 1–2 minutes.
Finish: Transfer to a plate and sprinkle with sesame seeds.

Notes

The magic is in the contrast — shatteringly crispy rice wrapped in a sticky, sweet-savory glaze with a gentle kick of heat. It works as a side dish, a snack, or honestly just dinner when no one's watching. 🍚🔥

Try this prompt

Mistral Large

Mistral Large

Mistral · Feb 2024text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

This matchup has more rounds

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