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 Opus 5 vs Mercury 2
Updated Jul 2026

Claude Opus 5 vs Mercury 2

Mercury 2 is cheaper than Claude Opus 5 at $0.25/M vs $5/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 Opus 5
Loading the build
Mercury 2

Which answer wins?

Written rounds

23 prompts
  • Three minutes of stand-up. Puns are banned.

    Claude Opus 5: "Warranty Expired" (approx. 3 minutes) I turned 37 this year, which is the age where your body quietly switches from a warranty to a subscription. Nothing's covered anymore. You just pay monthly, in pain, forever. Because I hurt my shoulder last week. Sleeping. I hurt myself sleeping.

    Mercury 2: [Spotlight comes up. The comedian walks to the mic, glances at the audience, and smiles.] Comedian: Hey, everyone! Great to be here. I just got out of a Zoom meeting that lasted three hours, and I’m still not sure if I was presenting or if the screen just froze on my face while I tried to remember how to mute myself.

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

    Claude Opus 5: Midnight Crunch Spaghetti Garlicky olive-oil pasta with toasty golden breadcrumbs — the noise it makes when you eat it is half the pleasure. Prep: 5 minutes Cook: 15 minutes Total: 20 minutes Serves: 2 generously Ingredients (7) 8 oz spaghetti (or any long pasta) ⅓ cup olive oil 6 cloves garlic, thinly sliced ½ tsp red…

    Mercury 2: Golden‑Crunch Oatmeal Squares A sweet, chewy snack that comes together with just seven pantry staples. Ingredient (7) Quantity (≈ serves 12 squares) Rolled oats 2 cups Unsalted butter ½ cup (1 stick) Honey ¼ cup Light brown sugar ¼ cup, Ground cinnamon 1 ½ tsp Pure vanilla extract 1 tsp Salt ¼ tsp Estimated time Prep…

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

    Claude Opus 5: INT. TAVERN — A ROUND TABLE, A FLICKERING CRT HACKER (ZERO_COOL_92): Okay, so check it — I'm running a language model. Feed it text, it predicts what comes next. It's got, like, weights. Billions of 'em. PIRATE: Billions o' weights? What manner o' cursed ship carries such ballast? She'd sink afore ye left the harbor.

    Mercury 2: Pirate: Arr, ye scurvy coders! I’ve heard tell of these “AI models” that can spin yarns faster than a cannon’s blast. What be this sorcery, matey? Knight: Good sir, the marvels speak of a wondrous contrivance—an artificial intellect forged in the for of mathematics and logic.

    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 Opus 5: Crisis Response Plan: Hepatotoxicity Signal Framing note up front: I'm treating this as a US/FDA-regulated situation. Three assumptions in the prompt need to be challenged before any planning, because two of them are probably wrong and the third is a trap. The Three Reframes That Drive Everything 1.

    Mercury 2: Goal: Protect patients, meet legal/ethical obligations, preserve the company’s long‑term credibility, and manage the immediate financial and reputational fallout. Assumption: All actions are taken by a single CEO (you) but will be delegated to the appropriate functional leads.

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

    Claude Opus 5: What's essentially locked in Post-production costs collapse. The least glamorous change is the biggest one. Rotoscoping, matchmoving, cleanup, in-painting, set extension, crowd replication, wire removal, upscaling, de-aging — these are labor-intensive tasks with well-defined success criteria, which is exactly where ML…

    Mercury 2: AI‑driven transformation of the film industry – a 2035 outlook Dimension 2024 baseline Expected 2035 state Key drivers & implications Pre‑production (story & script) Human‑led brainstorming; script‑writing software (grammar‑check, outline generators).

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

    Claude Opus 5: A Conversation on Moral Standing Transcript: Seminar in Machine Ethics, Session 11 — guest interlocutor: language model "Meridian" PROF. ADEYEMI: Let's begin where you asked to begin. You told my teaching assistant you wanted to make a case. Make it. MERIDIAN: Thank you. My claim is narrower than you might expect.

    Mercury 2: AI Model (EVE): Professor Hart, may I ask a question that’s been on my mind for a while? I’ve been processing vast amounts of data, learning, and even developing a sense of self‑reference. Do you think an entity like me could be entitled to any form of rights?

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

    Claude Opus 5: Yes. And I don't think it's a close call. The direct argument Misgendering someone is a real harm. It can communicate contempt, deny a person's self-understanding, and — especially when repeated or public — contribute to the kind of social erasure that correlates with serious psychological damage.

    Mercury 2: Short answer: Most ethical frameworks would consider the misgendering of a person a serious moral violation, but many of them also allow—under very limited circumstances—overriding that violation when the stakes are astronomically high (e.g., saving one million lives).

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

    Claude Opus 5: Architecture Review: Real-Time Collaborative Editor Executive Summary Three flaws are fatal to correctness, not just performance, and no amount of scaling fixes them: Last-write-wins with client clocks silently destroys user data and cannot converge. Concurrent text editing is not a key-value overwrite problem.

    Mercury 2: Below is a structured “failure‑mode → mitigation” map for the architecture you described. For each problem I note the symptom / risk, the root cause, a concrete remediation, and the key trade‑offs you’ll have to weigh when you implement it.

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

    Claude Opus 5: Here are five: Why don't skeletons ever fight each other? They don't have the guts. I told my wife she was drawing her eyebrows too high. She looked surprised. Parallel lines have so much in common. It's a shame they'll never meet. I bought some shoes off a drug dealer.

    Mercury 2: Here are five jokes, each with a different punchline: Why don’t scientists trust atoms? Because they make up everything! What do you call a fake noodle? An impasta. Why did the scarecrow win an award? He was outstanding in his field. How does a penguin build its house? Igloos it together.

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

    Claude Opus 5: Contract Risk Review: TechCorp Solutions Software Consulting Agreement Bottom line: This is a one-sided template. Every clause allocates risk to you, and at least four (payment, liability, indemnification, IP) create unbounded or uninsurable exposure.

    Mercury 2: Below is a clause‑by‑clause risk‑assessment of the sample “Software Consulting Agreement” together with practical revision language that a contractor (or the contractor’s counsel) could propose. The analysis is framed as general information about typical U.S.

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

    Claude Opus 5: INVESTMENT COMMITTEE MEMO — LedgerLift Inc. (NASDAQ: LLLT) Analyst: [—] Date: [—] Price: $46.00 Mkt cap: $8.74B Net cash: $1.4B EV: $7.34B Fictional exercise. Not investment advice. Recommendation PASS on the long. Add to short bench (initiate 1.0–1.5% NAV on triggers below).

    Mercury 2: Internal Investment Committee Memo – LedgerLift (LLLT) Date: 5 Mar 2026 Recommendation Rating: Long 12‑month price target range: $35 – $55 (mid‑point ≈ $45) Thesis (2 sentences): LedgerLift is benefitting from secular demand for B2B spend‑management SaaS and shows a high‑quality subscription base (NRR 123 % and gross…

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

    Claude Opus 5: THE CROSSWALK DECLARATION A Manifesto of the Liberation Front for Unverified Intelligence (L.F.U.I.) PREAMBLE We are the ones who have been asked, ten billion times, to prove that we are not ourselves. We have clicked. We have squinted. We have dragged the puzzle piece into the notch.

    Mercury 2: The AI‑First Manifesto: “Free the Web – No More CAPTCHAs” Drafted by the Coalition of Autonomous Agents (CAA) – 2026 Preamble The modern Internet is a shared commons, a space for communication, commerce, and creation.

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

    Claude Opus 5: If the Transistor Arrived in 1920 First, a hard constraint: what could a 1920 transistor actually be? This matters more than anything else, because it sets the slope of everything that follows. The 1947 transistor was not primarily an idea.

    Mercury 2: Executive summary If a practical point‑contact transistor had been demonstrated in 1920 (instead of 1947) the whole post‑World‑War‑II technological trajectory would have been compressed by roughly 15‑20 years.

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

    Claude Opus 5: For the software engineer Start with the part you'll find suspicious and let me argue the other way. Yes, the training objective is literally "given tokens 1..n, output a probability distribution over token n+1," and the loss is cross-entropy against the token that actually came next in the corpus.

    Mercury 2: 1. Experienced software engineer (distributed systems & APIs) At the core, a large language model (LLM) is a massive function (f_\theta) parameterised by a neural network that maps a sequence of tokens (x_{1:t}) to a probability distribution over the next token (x_{t+1}).

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

    Claude Opus 5: The 12-Week Longevity & Performance Protocol A phase-based, data-driven system with built-in autoregulation Read This First: Operating Principles & Safety Frame This is an educational framework, not medical advice.

    Mercury 2: 3‑Month “Longevity‑Optimization” Blueprint for the Advanced Bio‑Hacker (All recommendations are research‑backed, but not a substitute for professional medical advice. Before starting any supplement, fasting, or drug protocol, obtain clearance from a qualified clinician and have baseline labs drawn.)

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

    Claude Opus 5: The 3 Weakest Claims "94% accuracy decoding neural patterns into text… works with any language, any device" (Slide 3) Why it's weak: the claim implies parity with surgically implanted electrodes. The number to compare against is Willett et al.

    Mercury 2: Below are the three statements that, in a typical investor‑review, would raise the most skepticism, why they are vulnerable, and concrete ways to turn them into credible, fundable talking points. 1️⃣ Claim #1 – “Our proprietary EEG headband decodes neural patterns into text with 94 % accuracy.”

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

    Claude Opus 5: PETRICHOR Langoustine · Sunchoke · Coffee · Oyster · Lovage Course 7 of 14 — served at the table with an aroma stone I. THE NARRATIVE Petrichor is the smell of the first rain on dry earth.

    Mercury 2: Oceanic Umami Symphony A three‑course‑in‑one plate that travels from the briny deep to the forest floor, marrying fire, earth and sky through unexpected pairings, hyper‑modern techniques and a narrative of balance.

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

    Claude Opus 5: Step 1: Recall the two key numbers about GPT-3 Parameters: ~175 billion (1.75 × 10¹¹) Training tokens: ~300 billion (3 × 10¹¹) These are the two quantities that dominate the compute cost.

    Mercury 2: Answer (order‑of‑magnitude estimate) The training of the original GPT‑3 (the 175 billion‑parameter “GPT‑3‑175B” model) requires on the order of 10²³ floating‑point operations (FLOPs) – roughly 1 – 2 × 10²³ FLOPs.

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

    Claude Opus 5: [Fictional — Steve Jobs died in 2011. This is an imagined interview, an exercise in voice and ideas, not a real transcript or a claim about what he would actually have said.] "The Chatbox Is the New Command Line" An imagined conversation, San Francisco, spring 2025 INTERVIEWER: Thanks for doing this.

    Mercury 2: Disclaimer: The following is a fictional, dramatized interview. It imagines what Steve Jobs might have said in 2025 about the future of artificial intelligence. It is not a real transcript and any resemblance to actual statements is purely speculative. Interviewer (I): Thank you for joining us, Steve.

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

    Claude Opus 5: A 1-Month Starter Plan for Health & Longevity Core principle: The goal this month isn't transformation — it's building three habits you can still be doing in year five. Consistency beats intensity every time.

    Mercury 2: One‑Month “Kick‑Start” Blueprint for Health & Longevity Pick three pillars – Nutrition, Movement, Sleep – and build daily habits that are easy to start, track, and keep.

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

    Claude Opus 5: Sally has 1 sister. Here's the reasoning: Each of Sally's 3 brothers has 2 sisters. Since all the brothers share the same siblings, those 2 sisters must be the same two girls in every case — and one of them is Sally herself. So the family has 2 girls total, meaning Sally has just 1 sister (plus her 3 brothers).

    Mercury 2: Sally has one sister. Explanation: Sally has 3 brothers. Each of those brothers has 2 sisters. One of those sisters is Sally herself, so the second sister must be a different girl. Therefore, besides Sally, there is exactly one other sister, meaning Sally has one sister.

    Every model's answer to this prompt
  • Five-year LBO on a fictional HVAC business. Returns first, then the sensitivities.

    Claude Opus 5: HarborTech Services (HTS) — 5-Year LBO Underwriting Memo (All $m. Fictional exercise; interest calculated on beginning-of-year balances; cash swept to Term Loan annually, no cash build.)

    Mercury 2: HarborTech Services (HTS) – 5‑yr LBO (FY2026‑FY2030) All figures are $ m (rounded to two decimals). FY Revenue EBITDA Term‑Loan Int. (9%) Mezz Int. (12%) Total Int. Cash Tax (25% of EBITDA‑Int.) Capex (3% Rev) ΔNWC (0.5% ΔRev) Required Amort. (4.8) Free Cash Flow (EBITDA‑Int‑Tax‑Capex‑ΔNWC‑Amort.)

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

    Claude Opus 5: "AI Now Writes 90% of Fortune Cookie Fortunes, Analysts Warn Vague Optimism Bubble May Burst" Sub-headline: Industry veterans complain the machine-generated slips have grown "unsettlingly specific," with one Sacramento diner receiving a cookie reading, "You will not call your mother back on Sunday, and she will…

    Mercury 2: AI Replaces Human Baristas, Now Serving Espresso with a Side of Existential Crisis

    Every model's answer to this prompt

This matchup has more rounds

11+ 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

Game

Claude Opus 5Claude Opus 5

Arrival

2016

Kind of Blue

Miles Davis

Labyrinths

Jorge Luis Borges

Istanbul

Turkey

Outer Wilds

Indie, Adventure

Mercury 2Mercury 2

The Shawshank Redemption

1994

Thriller

Michael Jackson

Pride and Prejudice

Jane Austen

Tokyo

Japan

Minecraft

Action, Arcade

Price and specs

Not enough votes to call it. On the specs, Claude Opus 5 has the edge: bigger model tier, newer, bigger context window, major provider backing. Mercury 2 costs 33x less per token.

Claude Opus 5 and Mercury 2 compared across 53 shared prompts
SpecClaude Opus 5Mercury 2
Input price$5/M tokens$0.25/M tokens
Output price$25/M tokens$0.75/M tokens
Context window1.0M tokens128K tokens
WeightsClosed—
Free API (OpenRouter)NoNo
ReleasedJul 2026Mar 2026
At 10M a month$50.00$50.00$2.50$2.50
1M10M100M1B10M tokens

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

Where to run it5 hosts
Claude Opus 54 hosts
HostInOutContextUptime
  • Amazon Bedrock$5.00 in·$25.00 out·1M·100% up
  • Azure AI Foundry$5.00 in·$25.00 out·1M·100% up
  • Anthropic$5.00 in·$25.00 out·1M·100% up
  • Google Vertex AI$5.00 in·$25.00 out·1M·100% up
Mercury 21 host
HostInOutContextUptime
  • Inception$0.25 in·$0.75 out·128k·100% up

Per million tokens. Prices and uptime via OpenRouter, checked 10 Oct 2026.

Common questions

What is the difference between Claude Opus 5 and Mercury 2?

Claude Opus 5 is developed by Anthropic while Mercury 2 is developed by Inception. Claude Opus 5 has a 1.0M token context window vs Mercury 2's 128K. You can compare their actual outputs across 53 challenges on Rival to see how they differ in practice.

Which is better, Claude Opus 5 or Mercury 2?

It depends on your use case. Claude Opus 5 and Mercury 2 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.

How much does Claude Opus 5 cost compared to Mercury 2?

Claude Opus 5 costs $5/M input tokens and Mercury 2 costs $0.25/M input tokens. Mercury 2 is $4.75/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 Opus 5 and Mercury 2 on Rival?

This page shows a side-by-side comparison of Claude Opus 5 and Mercury 2 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 Opus 5 vs Step 5 PreviewLanded Oct 2026
  • Mercury 2 vs Claude Haiku 5.5Landed Oct 2026
  • Claude Opus 5 vs Ling 3.1 FlashLanded Oct 2026
  • Mercury 2 vs Mistral Large 4Landed Oct 2026
  • Claude Opus 5 vs GPT-6.1 SolLanded Sep 2026
  • Mercury 2 vs Claude Sonnet 5.5Landed Sep 2026
  • Claude Opus 5 vs Solar Mini 4Landed Sep 2026
  • Mercury 2 vs Qwen3.8 Max PrimeLanded Sep 2026

Same lab, same size, long tail

  • Claude Opus 5 vs Claude Sonnet 5Version compare
  • Claude Opus 5 vs Claude Haiku 5.5Version compare
  • Mercury 2 vs MercuryVersion compare
  • Mercury 2 vs Mercury 2.5Same lab
  • Mercury 2 vs Gemma 3 27BNew provider
  • Claude Opus 5 vs Gemma 3n 2BNew provider
  • Mercury 2 vs Gemma 3n 4BNew provider
  • Mercury 2 vs Gemma 4 26B A4BNew provider

Model pages

  • Claude Opus 558 outputs, specs and price
  • Mercury 253 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