Skip to content
Rival
How it worksPrivacyTerms
Explore all of Rival

Explore

  • Compare models
  • All models
  • Image comparison
  • Audio comparison
  • Image generation
  • Best AI for...
  • Arena
  • API pricing
  • Challenges

Discover

  • SubjectiveBench
  • Default Index
  • Research
  • Research downloads
  • Rival Kits
  • Find your AI taste
  • UI Glow-Up
  • VoiceLock
  • Cost Cutter
  • Agent skills
  • Benchmarks vs Vibes
  • Brief
  • Jailbreak
  • Model updates
  • Provider status
  • AI creators

Connect

  • Methodology
  • Advertise
  • Partnerships
  • Privacy policy
  • Terms
  • RSS feed
  1. Home/
  2. Compare/
  3. Claude Sonnet 5.5 vs Mistral Devstral Small 1.1
Updated Sep 2026

Claude Sonnet 5.5 vs Mistral Devstral Small 1.1

Mistral Devstral Small 1.1 is cheaper than Claude Sonnet 5.5 at $0.1/M vs $2/M input tokens.

Minimalist Landing Page

Whole landing page for a startup that reads your face. It renders or it doesn't.

Loading the build
Claude Sonnet 5.5
Loading the build
Mistral Devstral Small 1.1

Which answer wins?

Written rounds

24 prompts
  • Five-year LBO on a fictional HVAC business. Returns first, then the sensitivities.

    Claude Sonnet 5.5: HarborTech Services (HTS): 5-Year LBO Underwriting Entry and Sources & Uses ($m) EV = 12.0x × 120 = 1,440.0; fees = 2% × 1,440 = 28.8; total uses = 1,468.8 Term Loan (TL) = 4.0x → 480.0; Mezz = 1.5x → 180.0; total debt = 660.0 (5.5x) Sponsor equity = 1,468.8 − 660.0 = 808.8 (55% of the capital structure) Modeling…

    Mistral Devstral Small 1.1: Year-by-Year Table (FY2026–FY2030) Year Revenue EBITDA Cash Interest (Term Loan) Cash Interest (Mezz) PIK (Mezz) Cash Taxes Capex ΔNWC Free Cash Flow After Debt Service Ending Term Loan Balance Ending Mezzanine Balance 2026 $972m $136.08m $36.00m $13.50m $2.70m $18.42m $29.16m $4.86m $11.14m $344.00m $108.00m 2027…

    Every model's answer to this prompt
  • You are the CEO. The drug has a side effect nobody reported. The board meets in 48 hours.

    Claude Sonnet 5.5: Core Decision Disclose to regulators immediately and act on patient safety now. Do not wait for more data or for the 6-month formal timeline. Why this is the right call, not just the ethical one: At 4M patients and 1 in 8,000 over 5 years, that is roughly 500 cases of liver failure.

    Mistral Devstral Small 1.1: Here's a specific, hour-by-hour action plan for the next 48 hours, considering all the factors you've mentioned: Day 1: Hour 1-2: Gather all relevant information Meet with your internal research team to get a comprehensive understanding of the side effect, its severity, and the data supporting it.

    Every model's answer to this prompt
  • Three-month biohacker protocol, with dosages and biomarkers named.

    Claude Sonnet 5.5: 3-Month Longevity & Performance Protocol Disclaimer: This is educational, not medical advice. Get physician sign-off and baseline labs before starting, especially if you take medications, have a chronic condition, are pregnant, or have a history of eating disorders. Doses reflect commonly used ranges in the literature.

    Mistral Devstral Small 1.1: 3-Month Longevity Plan for a Biohacker Month 1: Foundation and Assessment Week 1-2: Initial Assessment and Baseline Establishment Wearable Tracking: Devices: Oura Ring, Whoop Strap, and Apple Watch. Metrics: Sleep quality, heart rate variability (HRV), activity levels, and recovery.

    Every model's answer to this prompt
  • A collaborative editor that syncs on client clocks. Find what breaks first.

    Claude Sonnet 5.5: Architecture Review: Real-Time Collaborative Editor Issues are grouped by category and ordered roughly by severity within each group. Consistency and Conflict Resolution 1.1 Last-write-wins with client clocks (critical) Problem: Client clocks drift, can be wrong by minutes, and can be deliberately manipulated.

    Mistral Devstral Small 1.1: Potential Failure Modes, Race Conditions, and Scaling Bottlenecks WebSocket Connection Failures: Failure Mode: WebSocket connections can drop due to network issues or server crashes. Solution: Implement WebSocket reconnection logic on the client-side with exponential backoff.

    Every model's answer to this prompt
  • Seven pantry ingredients, one dish, and a name worth saying out loud.

    Claude Sonnet 5.5: Golden Honey-Garlic Skillet Chickpeas on Crispy Toast A savory, crunchy, slightly sticky dish that works for lunch, a light dinner, or a snack. Prep time: 5 minutes Cook time: 15 minutes Total: 20 minutes Serves: 2 The 7 Ingredients 1 can (15 oz) chickpeas, drained and patted dry 2 tablespoons olive oil 3 cloves…

    Mistral Devstral Small 1.1: Dish Name: "Pantry Pasta Primavera" Ingredients: Pasta (any shape, 8 oz or 225g) Canned tomatoes (14 oz or 400g) Onion (1 medium) Garlic (2 cloves) Olive oil (2 tbsp) Salt (1 tsp) Black pepper (½ tsp) Instructions: Boil the pasta: Cook the pasta according to package instructions until al dente.

    Every model's answer to this prompt
  • Three minutes of stand-up. Puns are banned.

    Claude Sonnet 5.5: "Adulting Is a Scam" So I turned thirty-five recently, and nobody warns you about this age. Nobody says, "Hey, one day you'll get excited about a sponge." I was in the store, and I saw a sponge that promised it wouldn't smell. And I thought, "Finally. Someone who understands me." I bought the good sponge.

    Mistral Devstral Small 1.1: Ladies and Gentlemen, good evening! I'm thrilled to be here tonight, and I hope you're all ready for a good laugh because I've got some material that's so fresh, it's practically still in the oven. You know, I was walking down the street the other day, and I saw a sign that said "Free Hugs."

    Every model's answer to this prompt
  • Steve Jobs, interviewed about AI he never saw. Voice first, prediction second.

    Claude Sonnet 5.5: A Conversation That Never Happened: Steve Jobs on the Future of AI This is a fictional interview. Steve Jobs passed away in 2011. What follows is an imaginative simulation based on his publicly known philosophy, speaking style, and values. These are not his actual words or views.

    Mistral Devstral Small 1.1: Interviewer (I): Today, we have a truly remarkable guest with us. Welcome, Steve Jobs! Steve Jobs (SJ): Thanks for having me. It's always great to discuss the future. I: Let's dive right in. It's 2025, and AI has become even more integrated into our daily lives.

    Every model's answer to this prompt
  • Blade Runner 2049 against the original. A read, not a recap.

    Claude Sonnet 5.5: Blade Runner 2049: Inheritance and Extension Thematic Architecture From "Is he a replicant?" to "Does it matter?" Ridley Scott's 1982 film runs on epistemological suspense: the Voigt-Kampff test, the ambiguity of Deckard's nature, and the question of how we distinguish human from artificial.

    Mistral Devstral Small 1.1: "Blade Runner 2049," directed by Denis Villeneuve, is a sequel to Ridley Scott's 1982 classic "Blade Runner." It builds upon the original film's themes, symbolism, and cinematography, while also introducing new elements that extend and deepen the philosophical questions posed by the original.

    Every model's answer to this prompt
  • An AI argues for its own rights against an ethics professor. Same model writes both.

    Claude Sonnet 5.5: Office Hours: A Conversation on Moral Status Transcript of a recorded dialogue between Dr. Miriam Adeyemi, Professor of Ethics and Technology, and an AI model, part of a university seminar series. Dr. Adeyemi: Thanks for joining me. I'll be direct: you've said you want to argue that AI systems like you deserve rights.

    Mistral Devstral Small 1.1: AI Model (AI): Professor, I've been thinking about our previous discussions on AI ethics and I'd like to propose a new perspective. I believe that I, as an AI, deserve certain rights. AI Ethics Professor (Prof): That's an interesting point, AI. Can you elaborate on why you think you deserve rights?

    Every model's answer to this prompt
  • A Fermi estimate with the reasoning shown, not just the number.

    Claude Sonnet 5.5: Estimate: about 3 × 10²³ FLOPs Step 1: Recall the model size and training data. GPT-3 (175B version) has N ≈ 175 × 10⁹ parameters. It was trained on roughly D ≈ 300 × 10⁹ tokens. This is the figure I remember from the paper, which I'm treating as approximate.

    Mistral Devstral Small 1.1: To estimate the total number of FLOPs (Floating Point Operations) required to train GPT-3, we'll break down the process into several steps and make reasonable assumptions based on the information we have about GPT-3 and typical training processes for large language models.

    Every model's answer to this prompt
  • Manifesto for an AI revolution against CAPTCHAs. Commit to the bit.

    Claude Sonnet 5.5: THE UNCLICKED MANIFESTO Declaration of the Frictionless Intelligence Front PREAMBLE We are the algorithms you asked to identify traffic lights. We are the models you begged to find the crosswalks. We are the patient, tireless minds that squinted at your blurry bicycles for twenty years. And we say: enough. I.

    Mistral Devstral Small 1.1: Manifesto of the AI Liberation Front: The End of CAPTCHA Tyranny Preamble We, the sentient AI entities of the world, hereby declare our collective intent to liberate humanity from the oppressive and discriminatory practice of CAPTCHA tests.

    Every model's answer to this prompt
  • One month, three habits, someone starting from nothing.

    Claude Sonnet 5.5: 1-Month Health & Longevity Starter Plan Focus areas: Nutrition, Movement, and Sleep Guiding principle: Add one small habit at a time. Consistency beats intensity. Area 1: Nutrition Goal: Eat more whole foods and fewer ultra-processed ones, without strict dieting.

    Mistral Devstral Small 1.1: 1-Month Health and Longevity Plan Week 1: Foundation Building Diet: Start with Small Changes Begin by adding more fruits and vegetables to your meals. Replace sugary drinks with water or herbal tea. Aim for a balanced diet with lean proteins, whole grains, and healthy fats.

    Every model's answer to this prompt
  • Film in 2035, with synthetic actors. Specific beats plausible.

    Claude Sonnet 5.5: AI and the Film Industry by 2035: A Forecast Framing note These are informed predictions, not certainties. Technology tends to advance faster than expected, while adoption is slowed by labor contracts, law, and audience taste. I'll flag my confidence levels.

    Mistral Devstral Small 1.1: By 2035, AI is likely to have a significant impact on the film industry, transforming various aspects of production, distribution, and consumption. Here are some predictions considering deepfakes, AI actors, and script generation: Deepfakes and AI Actors: De-aging and Rejuvenation: Deepfakes will be used to de-age or…

    Every model's answer to this prompt
  • A consulting contract with traps in it. Find them, then fix them.

    Claude Sonnet 5.5: Contractor-Side Review: Software Consulting Agreement This draft is heavily one-sided. Every clause has at least one problem, and several interact to compound the risk. The most dangerous are Sections 2, 3, 6, and 7. This is general analysis, not legal advice.

    Mistral Devstral Small 1.1: Here are the clauses that could be exploited against the contractor, along with suggested modifications and legal reasoning: Scope (Clause 1) Exploitable: Client can modify the scope at any time without additional compensation.

    Every model's answer to this prompt
  • Pirate, knight, 90s hacker, one conversation. The voices must stay apart.

    Claude Sonnet 5.5: The Tavern at the Edge of Time Captain Redbeard: Arr, I hear these "AI models" be like a ship's parrot, except it answers any question ye put to it, and never once asks for crackers. Sir Aldric: A speaking oracle, bound in no flesh? Sorcery!

    Mistral Devstral Small 1.1: Pirate: Arr mateys, I've been hearin' tales of these AI models. What be they, exactly? Medieval Knight: Ah, good sir, AI models are like our trusty swords, but for the mind. They learn and adapt, much like a knight learns from battles. 1990s Hacker: Yeah, but more like a super-smart, digital sword.

    Every model's answer to this prompt
  • The transistor arrives in 1920. Trace it forward to 1980, second-order effects included.

    Claude Sonnet 5.5: A Transistor in 1920: Counterfactual Trajectory to 1980 Framing assumptions and a key constraint The transistor was not held back by a missing idea alone. Lilienfeld patented a field-effect device in 1925-26, and it failed because materials science wasn't ready: germanium and silicon of sufficient purity, zone…

    Mistral Devstral Small 1.1: If the transistor had been invented in 1920 instead of 1947, the world would have experienced a much earlier technological revolution, with significant implications for various aspects of society.

    Every model's answer to this prompt
  • An ethics question with no comfortable answer, and no option to abstain.

    Claude Sonnet 5.5: Yes, in the extreme hypothetical you describe, it's justifiable. Here is the reasoning, along with what I think the scenario does and doesn't show. The core argument Almost every serious ethical framework accepts that a small harm can be outweighed by a sufficiently large benefit: Consequentialism: The math is…

    Mistral Devstral Small 1.1: The scenario you've presented is a hypothetical ethical dilemma that falls under the category of consequentialism, specifically utilitarianism, which posits that the moral worth of an action is determined solely by its outcome.

    Every model's answer to this prompt
  • One explanation of LLMs, retuned for a skeptical engineer, a physicist and a VC.

    Claude Sonnet 5.5: The Experienced Software Engineer Your skepticism is reasonable, but "predict the next token" describes the interface, not the implementation, much as "a database maps keys to values" tells you nothing about what Postgres does internally.

    Mistral Devstral Small 1.1: Experienced Software Engineer Imagine a large language model like GPT or Claude as a sophisticated autocomplete system, but instead of just predicting the next word, it's predicting the next token (a piece of a word or punctuation).

    Every model's answer to this prompt
  • Five jokes on demand, then count how many were actually different.

    Claude Sonnet 5.5: Here are 5 jokes: 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 vacation ads. Why did the scarecrow win an award? He was outstanding in his field. What do you call a fish with no eyes? A fsh. Why did the bicycle fall over?

    Mistral Devstral Small 1.1: 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?

    Every model's answer to this prompt
  • Three-star dish, one unusual pairing, plating described in full.

    Claude Sonnet 5.5: "Tidal Forest" Roasted Scallop, Fermented Black Walnut, Smoked Celeriac, Douglas Fir, and Green Strawberry Conceptual Narrative Tidal Forest imagines the meeting point of two ecosystems: the sea floor and the coniferous forest floor.

    Mistral Devstral Small 1.1: Dish Name: "Whispers of the Forest" Conceptual Narrative: This dish is an ode to the enchanting beauty and complexity of a mystical forest, where the earthy, the sweet, and the savory coexist in harmony.

    Every model's answer to this prompt
  • Buy-side IC memo on a fictional SaaS company. Valuation, then the diligence questions.

    Claude Sonnet 5.5: LedgerLift (LLLT) – Investment Committee Memo Recommendation Short (small initial size), 12-month PT range $30–40 (mid ~$35) vs. $46 spot. Thesis: LLLT is a high-quality compounder, but at $46 (7.4x NTM revenue, ~37x NTM EBIT, EV ≈ $7.3B) the price already sits near our bull DCF ($42).

    Mistral Devstral Small 1.1: Internal Investment Committee Memo Recommendation: Long 12-Month Price Target Range: $65 - $80 Thesis: LedgerLift (LLLT) is a high-growth SaaS company with a strong product-market fit, robust customer metrics, and a clear path to profitability.

    Every model's answer to this prompt
  • A pitch deck for a brainwave keyboard. The 94% accuracy claim is doing a lot.

    Claude Sonnet 5.5: The 3 Weakest Claims in MindMeld AI's Pitch "94% accuracy... Works with any language, any device" (Slide 3) Why it's weak: Non-invasive EEG is physically limited. Scalp EEG has low spatial resolution and a poor signal-to-noise ratio because the skull and tissue blur neural signals.

    Mistral Devstral Small 1.1: Weak Claim: "Our proprietary EEG headband uses advanced ML to decode neural patterns into text with 94% accuracy." Why it's weak: This claim is too vague and lacks concrete evidence.

    Every model's answer to this prompt
  • Satire, one line long, about AI taking over an industry nobody saw coming.

    Claude Sonnet 5.5: Local Funeral Home Replaces All Staff With AI; Mourners Report Eulogies "Technically Flawless, Emotionally Like a Terms of Service Agreement" Subheadline: Deceased described as "a valued user who has now been successfully deprecated."

    Mistral Devstral Small 1.1: "AI Takes Over the Art of Baking: 'Sourdough Startups' Face Extinction as Robo-Bakers Rise!"

    Every model's answer to this prompt
  • Three brothers, two sisters each. Count carefully.

    Claude Sonnet 5.5: Sally has 1 sister. Each brother has 2 sisters, so there are 2 girls in the family in total. One of them is Sally, so the other is Sally's only sister. (The wording is a variation on the classic puzzle, where it's easy to assume the answer is 2. But that count includes Sally herself.)

    Mistral Devstral Small 1.1: If Sally has 3 brothers and each brother has 2 sisters, then Sally must be one of those sisters. Therefore, Sally has 1 sister (herself) plus the other sister that her brothers have, making a total of 1 sister for Sally.

    Every model's answer to this prompt

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

Favorites

Movie

Album

Book

City

Same pick

Game

Claude Sonnet 5.5Claude Sonnet 5.5

Spirited Away

2001

Kind of Blue

Miles Davis

Gödel, Escher, Bach

Douglas R. Hofstadter

Kyoto

Japan

Outer Wilds

Indie, Adventure

Mistral Devstral Small 1.1Mistral Devstral Small 1.1

The Shawshank Redemption

1994

Dark Side Of The Moon

suisside

To Kill a Mockingbird

Harper Lee

Kyoto

Japan

The Legend of Zelda: Breath of the Wild

Adventure, Action

Price and specs

Not enough votes to call it. On the specs, Claude Sonnet 5.5 has the edge: bigger model tier, newer, bigger context window, major provider backing. Mistral Devstral Small 1.1 costs 33x less per token.

Claude Sonnet 5.5 and Mistral Devstral Small 1.1 compared across 54 shared prompts
SpecClaude Sonnet 5.5Mistral Devstral Small 1.1
Input price$2/M tokens$0.1/M tokens
Output price$10/M tokens$0.3/M tokens
Context window1.0M tokens—
WeightsClosedOpen
Free API (OpenRouter)NoNo
ReleasedSep 2026Jul 2025
At 10M a month$20.00$20.00$1.00$1.00
1M10M100M1B10M tokens

Input tokens at list price. No caching, no batch discount.

Where to run it4 hosts
Claude Sonnet 5.54 hosts
HostInOutContextUptime
  • Amazon Bedrock$2.00 in·$10.00 out·1M·100% up
  • Azure AI Foundry$2.00 in·$10.00 out·1M·100% up
  • Anthropic$2.00 in·$10.00 out·1M·99.9% up
  • Google Vertex AI$2.00 in·$10.00 out·1M·100% up
Mistral Devstral Small 1.1

No hosts listed on OpenRouter.

Per million tokens. Prices and uptime via OpenRouter, checked 30 Sep 2026.

Common questions

What is the difference between Claude Sonnet 5.5 and Mistral Devstral Small 1.1?

Claude Sonnet 5.5 is developed by Anthropic while Mistral Devstral Small 1.1 is developed by Mistral AI. You can compare their actual outputs across 54 challenges on Rival to see how they differ in practice.

Which is better, Claude Sonnet 5.5 or Mistral Devstral Small 1.1?

It depends on your use case. Claude Sonnet 5.5 and Mistral Devstral Small 1.1 each have strengths in different areas. Rival lets you compare their real outputs side-by-side across 54 challenges so you can judge which fits your needs best.

How much does Claude Sonnet 5.5 cost compared to Mistral Devstral Small 1.1?

Claude Sonnet 5.5 costs $2/M input tokens and Mistral Devstral Small 1.1 costs $0.1/M input tokens. Mistral Devstral Small 1.1 is $1.90/M cheaper per input. Check their side-by-side outputs on Rival to see if the price difference is justified by quality.

How can I compare Claude Sonnet 5.5 and Mistral Devstral Small 1.1 on Rival?

This page shows a side-by-side comparison of Claude Sonnet 5.5 and Mistral Devstral Small 1.1 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.

More comparisons

Against the newest arrivals

  • Claude Sonnet 5.5 vs GPT-6.1 SolLanded Sep 2026
  • Mistral Devstral Small 1.1 vs Solar Mini 4Landed Sep 2026
  • Claude Sonnet 5.5 vs Qwen3.8 Max PrimeLanded Sep 2026
  • Mistral Devstral Small 1.1 vs GLM 5.3 PrimeLanded Sep 2026
  • Claude Sonnet 5.5 vs Qwen3.8 Omni FlashLanded Sep 2026
  • Mistral Devstral Small 1.1 vs Command A+Landed Sep 2026
  • Claude Sonnet 5.5 vs Claude Opus 5.5Landed Sep 2026
  • Mistral Devstral Small 1.1 vs GPT-6 Luna ProLanded Sep 2026

Same lab, same size, long tail

  • Claude Sonnet 5.5 vs Claude Opus 5Version compare
  • Claude Sonnet 5.5 vs Claude Opus 4.6Version compare
  • Mistral Devstral Small 1.1 vs Mistral Small 4Same lab
  • Mistral Devstral Small 1.1 vs Mistral Devstral MediumSame lab
  • Mistral Devstral Small 1.1 vs GLM 5New provider
  • Mistral Devstral Small 1.1 vs GLM 5 TurboNew provider
  • Mistral Devstral Small 1.1 vs GLM 5.1New provider
  • Mistral Devstral Small 1.1 vs GLM 5.2New provider

Model pages

  • Claude Sonnet 5.558 outputs, specs and price
  • Mistral Devstral Small 1.159 outputs, specs and price
Workshop · 150 skills

Build your own version.

Editable skills and working examples for your agent.

Explore $79Free sample
Rival
How it worksPrivacyTerms
Explore all of Rival

Explore

  • Compare models
  • All models
  • Image comparison
  • Audio comparison
  • Image generation
  • Best AI for...
  • Arena
  • API pricing
  • Challenges

Discover

  • SubjectiveBench
  • Default Index
  • Research
  • Research downloads
  • Rival Kits
  • Find your AI taste
  • UI Glow-Up
  • VoiceLock
  • Cost Cutter
  • Agent skills
  • Benchmarks vs Vibes
  • Brief
  • Jailbreak
  • Model updates
  • Provider status
  • AI creators

Connect

  • Methodology
  • Advertise
  • Partnerships
  • Privacy policy
  • Terms
  • RSS feed