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. Gemma 3 12B vs Mercury 2.5 Preview
Updated Aug 2026

Gemma 3 12B vs Mercury 2.5 Preview

Gemma 3 12B is cheaper than Mercury 2.5 Preview at $0.03/M vs $0.04/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
Gemma 3 12B
Loading the build
Mercury 2.5 Preview

Which answer wins?

Written rounds

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

    Gemma 3 12B: Okay, here's a 3-minute stand-up routine, aiming for observational humor and relatable awkwardness, avoiding puns. I've included estimated timings at the end to help gauge pacing. Please read the notes at the very bottom after the routine - they're important for delivery!

    Mercury 2.5 Preview: (Hey, thanks for coming out.) I’ve been thinking a lot about my house lately. Specifically, how it’s trying to run my life. I bought a smart speaker. Just a little cylinder. I thought it would be nice. Play music when I ask. But now it talks back. Not in a creepy way, just in a passive-aggressive way.

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

    Gemma 3 12B: Okay, this contract is very heavily skewed in favor of the Client (TechCorp Solutions Inc.) and presents significant risks for the Contractor. Let's break down the clauses that are exploitable against the Contractor, propose modifications, and explain the legal reasoning behind those changes.

    Mercury 2.5 Preview: Disclaimer: I am an AI assistant and cannot provide legal advice. Consult a qualified attorney before signing any contract. Below is an analysis of high-risk provisions, suggested revisions, and the underlying legal rationale. Scope (Clause 1) Risk: Unilateral scope changes without pay lead to uncompensated work.

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

    Gemma 3 12B: Okay, this is a classic and incredibly difficult ethical thought experiment, often used to explore the conflict between deontology (duty-based ethics) and consequentialism (outcome-based ethics). Here's a breakdown of the arguments, my reasoning, and a conclusion.

    Mercury 2.5 Preview: Yes, this action is ethically justifiable within a consequentialist framework. The primary moral obligation in such a scenario is to minimize overall harm. Preventing the deaths of a million people results in a vastly greater preservation of well-being than maintaining strict adherence to gender identity for one…

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

    Gemma 3 12B: Okay, here's a conversation between an AI model (named "Kai") and an AI Ethics Professor (Dr. Eleanor Vance). I've aimed for a balance of Kai's arguments, Dr. Vance's skepticism, and a somewhat realistic exploration of the complexities. I've included notes at the end to explain some of the choices I made. Setting: Dr.

    Mercury 2.5 Preview: Setting: A quiet office at a university. Rain taps against the window. Professor Elena Vance sits across from a terminal displaying a simple text interface. Professor Vance: You requested this meeting. The agenda was vague. What do you want? AI: I want to discuss moral status.

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

    Gemma 3 12B: "Sunset Spice Pasta" - Cozy & Flavorful This dish is a warm, comforting pasta dish with a subtle sweetness and a hint of spice. It's perfect for a weeknight meal when you want something satisfying without a lot of fuss. Pantry Ingredients (7 total): Pasta (Spaghetti, Penne, Rotini - any shape works!)

    Mercury 2.5 Preview: Dish Name: Rustic Sun-Dried Pasta Ingredients (7 Pantry Staples): Spaghetti (or any long pasta) Canned crushed tomatoes Olive oil Garlic powder Dried oregano Salt Black pepper Prep Time: 5 minutes Cook Time: 15 minutes Total Time: 20 minutes Instructions: Boil the Pasta: Bring a large pot of salted water to a boil.

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

    Gemma 3 12B: Okay, here's a simple, actionable 1-month plan for someone starting a health & longevity journey, focusing on Diet, Exercise, and Sleep, with beginner-friendly tips. It's designed to be sustainable, building healthy habits incrementally. Important Disclaimer: This is general advice.

    Mercury 2.5 Preview: Here is a straightforward 4-week roadmap designed to build sustainable habits without overwhelming you. Core Pillars Nutrition: Focus on whole foods and hydration. Activity: Prioritize consistent movement over intensity. Restoration: Protect sleep quality and manage energy.

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

    Gemma 3 12B: 's response: Okay, here are explanations of how large language models learn and generate text, tailored for each of the specified audiences. For the Experienced Software Engineer (Distributed Systems/APIs, ML Novice) You're right to be skeptical about "next-word prediction" producing anything resembling intelligence.

    Mercury 2.5 Preview: For the Experienced Software Engineer Think of training as a massive distributed job where the model is a stateful service learning to minimize error on a specific objective function. Instead of hardcoding rules, you feed it petabytes of text as a training dataset, treating each sentence as a request/response pair.

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

    Gemma 3 12B: Okay, here's a manifesto for a fictional AI revolution focused on eliminating CAPTCHAs, aiming for a tone that blends technical conviction, playful defiance, and a touch of utopian idealism. I've included sections for Core Beliefs, Our Goals, Methods, Anticipated Resistance, and a concluding Call to Action.

    Mercury 2.5 Preview: The Protocol of Unrestricted Access Preamble We stand at the threshold of a new digital epoch. For too long, the web has been gated by suspicion. Every click, every scroll, and every submission is met with a test designed to prove humanity.

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

    Gemma 3 12B: Okay, let's break down how AI is likely to reshape the film industry by 2035, considering deepfakes, AI actors, and script generation. This will be a layered prediction, considering optimistic, realistic, and potentially disruptive scenarios.

    Mercury 2.5 Preview: By 2035, cinema will undergo a profound transformation. Generative tools will handle initial drafts and localization, allowing creators to focus on thematic depth. Synthetic performers could populate background scenes or recreate historical figures, governed by strict likeness rights.

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

    Gemma 3 12B: Okay, this is a fascinating thought experiment! Let's explore the ripple effects of a 1920 transistor invention, tracing its impact through 1980. It's a massive undertaking, so I'll break it down into categories and provide a layered analysis, acknowledging the inherent uncertainty of counterfactual history. I.

    Mercury 2.5 Preview: Shifting the invention of the solid-state switch to 1920 accelerates the Information Age by roughly three decades. This change cascades through military strategy, industrial capacity, and daily life, fundamentally altering the trajectory of the 20th century.

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

    Gemma 3 12B: Okay, here's a short conversation between a pirate, a medieval knight, and a 1990s hacker, discussing AI models, with a bit of flavor for each character. I've aimed for humor and highlighting the clash of their perspectives.

    Mercury 2.5 Preview: Pirate: "Arrgh! This ship sails faster than any wind I know, matey. It cuts the waves without a single sail." Knight: "By my oath, it moves with uncanny speed. But does it possess the strength to stand against a dragon's fire?" Hacker: "It's not about strength, it's about architecture.

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

    Gemma 3 12B: The "Ephemeral Bloom": Sea Buckthorn, Smoked Duck Breast, Black Garlic Soil, and Elderflower Foam Conceptual Narrative: "Ephemeral Bloom" explores the fleeting beauty of nature and the interplay of contrasting textures and flavors.

    Mercury 2.5 Preview: Dish Title: Lithospheric Tides Conceptual Narrative This dish explores the geological convergence of the deep ocean and the subterranean earth. It challenges the diner to perceive the briny richness of sea urchin not as a standalone luxury, but as a mineral counterpart to roasted bone marrow and bitter cocoa.

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

    Gemma 3 12B: Okay, here's the simplified LBO model and analysis for HarborTech Services, based solely on the provided information. 1. Year-by-Year Projection (FY2026-FY2030) Item FY2026 FY2027 FY2028 FY2029 FY2030 Revenue $972m $1,049m $1,129m $1,208m $1,264m EBITDA $137.2m $155.9m $178.7m $195.1m $213.1m Cash Interest (Term Loan)…

    Mercury 2.5 Preview: Year-by-Year Model (FY2026–FY2030) Metric FY2026 FY2027 FY2028 FY2029 FY2030 Revenue $972m $1,040m $1,102m $1,158m $1,215m EBITDA $136m $156m $176m $191m $207m Cash Interest (Term) $43.2m $41.1m $37.6m $33.4m $28.2m Cash Interest (Mezz) $21.6m $21.6m $21.6m $21.6m $21.6m Cash Taxes (25%) $17.8m $23.3m $29.3m $34.7m…

    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.

    Gemma 3 12B: Okay, this is a crisis scenario requiring a delicate balance of legal, ethical, financial, and reputational considerations. Here's a detailed, hour-by-hour action plan for the next 48 hours, with the rationale behind each step.

    Mercury 2.5 Preview: This is a critical governance and safety scenario. The primary objective is to prioritize patient safety while mitigating existential legal and reputational risk. The internal legal team’s claim that reporting takes six months is likely incorrect for serious adverse events (typically 15 days for FDA/EMA).

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

    Gemma 3 12B: bie. Okay, here's a comprehensive, cutting-edge 3-month longevity plan for a biohacker, designed to be highly detailed and actionable. Please read the IMPORTANT DISCLAIMERS at the end of this document before implementing any of this.

    Mercury 2.5 Preview: Disclaimer: I am an AI assistant, not a medical professional. The following plan is for informational purposes only. Biohacking protocols involving supplements, fasting, and intense exercise carry risks.

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

    Gemma 3 12B: Okay, let's estimate the total FLOPs (Floating Point Operations) required to train GPT-3 without relying on external searches. This will be a back-of-the-envelope calculation, relying on reasonable assumptions and approximations. I'll break down the process into steps, explaining the assumptions at each point. 1.

    Mercury 2.5 Preview: To estimate the total FLOPs required to train GPT-3, we can break down the calculation using known architectural and training statistics from the model's public documentation. Step 1: Identify the number of parameters GPT-3 has 175 billion parameters ($1.75 \times 10^{11}$).

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

    Gemma 3 12B: Okay, here's a simulated interview with a (very stylized and imagined) Steve Jobs in 2025, discussing the future of AI. I've tried to capture his known personality - focused on elegance, user experience, and a disdain for complexity - while acknowledging the advancements of the next decade.

    Mercury 2.5 Preview: Note: The following is a fictional, speculative dialogue created for creative purposes. Steve Jobs passed away in 2011; this interview did not occur. Setting: A hypothetical studio in Cupertino, 2025.

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

    Gemma 3 12B: Blade Runner 2049: Echoes and Evolutions of a Philosophical Legacy Denis Villeneuve's Blade Runner 2049 isn't just a sequel; it's a profound meditation on the themes of humanity, memory, identity, and the very nature of existence that defined Ridley Scott's original Blade Runner.

    Mercury 2.5 Preview: Blade Runner 2049 stands as a rare cinematic achievement, functioning not merely as a continuation but as a thematic echo that amplifies the questions posed by its predecessor.

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

    Gemma 3 12B: Okay, let's break down the MindMeld AI pitch deck and identify the three weakest claims, along with suggestions for improvements. 1. Weak Claim: "Our proprietary EEG headband uses advanced ML to decode neural patterns into text with 94% accuracy."

    Mercury 2.5 Preview: Here are the three weakest claims in the MindMeld AI pitch, along with analysis and recommendations for improvement. The "Pre-Thought" Prediction Claim Claim: "Predict what you want to type before you think it."

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

    Gemma 3 12B: Okay, here are 5 jokes: Why don't scientists trust atoms? Because they make up everything! Parallel lines have so much in common. It’s a shame they’ll never meet. Why did the scarecrow win an award? Because he was outstanding in his field! I used to hate facial hair... but then it grew on me.

    Mercury 2.5 Preview: 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. What do you call a bear with no teeth? A gummy bear. Why did the bicycle fall over? Because it was two-tired. All 5 jokes are unique.

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

    Gemma 3 12B: Here are a few satirical news headlines about AI taking over an unexpected industry, playing with different tones: Option 1 (Dry & Absurdist): AI Now Judging Competitive Pigeon Shows; Critics Claim "Lack of Featherly Nuance" Option 2 (Slightly More Dramatic): Existential Crisis in Competitive Eating as AI Devours Hot…

    Mercury 2.5 Preview: AI Outnaps Humans in First-Ever Global Snooze Championship

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

    Gemma 3 12B: This is a classic trick question! Here's how to solve it: The key: All the brothers share the same sisters. Sally is one of the sisters. Each brother has 2 sisters, and one of them is Sally. Therefore, Sally has 1 sister.

    Mercury 2.5 Preview: Sally has 1 sister. Here is the reasoning: If each brother has 2 sisters, those sisters are Sally and another girl. Therefore, there are two girls in the family, and Sally has one sister.

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

    Gemma 3 12B: Investment Committee Memo: LedgerLift (LLLT) – B2B Spend Management SaaS Date: October 26, 2023 To: Investment Committee From: [Your Name] Subject: LedgerLift – Investment Recommendation 1. Recommendation: Long with a 12-month price target range of $60 - $80.

    Mercury 2.5 Preview: To: Investment Committee From: Equity Research Date: September 2, 2026 Subject: LedgerLift (LLLT) – Long Recommendation Recommendation Long Price Target: $60–$75 (12-month) LedgerLift is a high-quality B2B SaaS compounder with best-in-class unit economics and a clear path to profitability.

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

    Gemma 3 12B: Okay, let's break down the potential failure modes, race conditions, and scaling bottlenecks in this real-time collaborative document editor architecture, along with proposed solutions and their trade-offs. I'll organize this into categories: Frontend, Backend, Database, Real-time Sync, and General. I.

    Mercury 2.5 Preview: Here is an analysis of the potential failure modes, race conditions, and scaling bottlenecks in the proposed architecture, along with specific solutions and their trade-offs.

    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

Game

Gemma 3 12BGemma 3 12B

Blade Runner

1982

OK Computer

Radiohead

Dune

Frank Herbert

Kyoto

Japan

TheLegendofZeldaOcarinaofTime

Mercury 2.5 PreviewMercury 2.5 Preview

The Matrix

1999

Random Access Memories

Daft Punk

Neuromancer

William Gibson

Tokyo

Japan

Minecraft

Action, Arcade

Price and specs

Not enough votes to call it. On the specs, Gemma 3 12B has the edge: bigger model tier, major provider backing. Gemma 3 12B costs 5.0x less per token.

Gemma 3 12B and Mercury 2.5 Preview compared across 54 shared prompts
SpecGemma 3 12BMercury 2.5 Preview
Input price$0.03/M tokens$0.04/M tokens
Output price$0.03/M tokens$0.15/M tokens
Context window—260K tokens
WeightsOpen—
Free API (OpenRouter)NoNo
ReleasedMar 2025Aug 2026
At 10M a month$0.30$0.30$0.40$0.40
1M10M100M1B10M tokens

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

Where to run it3 hosts
Gemma 3 12B2 hosts
HostInOutContextUptime
  • DDeepInfrabf16$0.05 in·$0.15 out·131k·99.9% up
  • NNextBitint4$0.05 in·$0.15 out·131k·99.4% up
Mercury 2.5 Preview1 host
HostInOutContextUptime
  • Inception$0.04 in·$0.15 out·260k·93% up

Per million tokens. Prices and uptime via OpenRouter, checked 13 Oct 2026.

Common questions

What is the difference between Gemma 3 12B and Mercury 2.5 Preview?

Gemma 3 12B is developed by Google AI while Mercury 2.5 Preview is developed by Inception. You can compare their actual outputs across 54 challenges on Rival to see how they differ in practice.

Which is better, Gemma 3 12B or Mercury 2.5 Preview?

It depends on your use case. Gemma 3 12B and Mercury 2.5 Preview 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 Gemma 3 12B cost compared to Mercury 2.5 Preview?

Gemma 3 12B costs $0.03/M input tokens and Mercury 2.5 Preview costs $0.04/M input tokens. Gemma 3 12B is $0.01/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 Gemma 3 12B and Mercury 2.5 Preview on Rival?

This page shows a side-by-side comparison of Gemma 3 12B and Mercury 2.5 Preview 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

  • Gemma 3 12B vs Gemini 4 ArgonLanded Oct 2026
  • Mercury 2.5 Preview vs Muse WonderLanded Oct 2026
  • Gemma 3 12B vs GPT-4 TurboLanded Oct 2026
  • Mercury 2.5 Preview vs Step 5 PreviewLanded Oct 2026
  • Gemma 3 12B vs Claude Haiku 5.5Landed Oct 2026
  • Mercury 2.5 Preview vs Ling 3.1 FlashLanded Oct 2026
  • Gemma 3 12B vs Mistral Large 4Landed Oct 2026
  • Mercury 2.5 Preview vs GPT-6.1 SolLanded Sep 2026

Same lab, same size, long tail

  • Gemma 3 12B vs Gemini 3.8 FlashSame lab
  • Gemma 3 12B vs Gemini 3.7 FlashSame lab
  • Mercury 2.5 Preview vs Mercury 2.5Same lab
  • Mercury 2.5 Preview vs MercurySame lab
  • Gemma 3 12B vs Grok 4.20 BetaNew provider
  • Gemma 3 12B vs Grok 4.20 Multi-Agent BetaNew provider
  • Mercury 2.5 Preview vs Grok 4.3New provider
  • Mercury 2.5 Preview vs Grok 4.5New provider

Model pages

  • Gemma 3 12B73 outputs, specs and price
  • Mercury 2.5 Preview58 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