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. GPT-5.6 Terra vs MiniMax M2-her
Updated Jul 2026

GPT-5.6 Terra vs MiniMax M2-her

MiniMax M2-her is cheaper than GPT-5.6 Terra at $0.3/M vs $2.5/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
GPT-5.6 Terra
Loading the build
MiniMax M2-her

Which answer wins?

Written rounds

24 prompts
  • Three-star dish, one unusual pairing, plating described in full.

    GPT-5.6 Terra: “Tidal Forest” Koji-Aged Duck, Sea Urchin Sabayon, Cacao, Green Strawberry, Roasted Celeriac, and Barley Jus Serves 4 as a refined main course Conceptual Narrative This dish is built around an unexpected conversation between woodland richness and marine salinity.

    MiniMax M2-her: Alright, I can certainly help craft an innovative recipe for a multi-component dish worthy of a Michelin 3-star restaurant. Let me take you through the process step by step, just like a chef would. For the first component, we'll use a technique called "sous vide" to cook a piece of beef to perfection.

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

    GPT-5.6 Terra: Fictional, imagined interview — not real quotes. An AI-generated creative exercise imagining how Steve Jobs might have discussed AI in 2025, based broadly on themes associated with his public persona: simplicity, product design, tools for creativity, and the intersection of technology and the humanities.

    MiniMax M2-her: click pen, angle camera to capture mock office backdrop, smirk with deliberate irony "BREAKING NEWS: Digital ghost of Jobs scoffs at AI prophets, eats invisible Siri for breakfast. Switched his ghost to Android just to trigger Tim Cook in the afterlife. REAL headline. Now send $19.99 to revive his hologram."

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

    GPT-5.6 Terra: Pirate: Arrr, this AI model be a strange beast. Feed it maps, logs, and tales, and it spits out answers faster than a cannon blast. Knight: Verily, but one must not trust every proclamation it makes. A model may speak with confidence yet possess no true wisdom. Hacker: Yeah, it’s basically pattern prediction at scale.

    MiniMax M2-her: Pirate: Arr, me hearties! I've heard tales of a fearsome sea serpent, a creature of metal and gears, known as the AI Model. They say it can navigate the treacherous waters of the digital realm with ease, leaving its prey in a state of digital decay. What say ye, me mateys?

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

    GPT-5.6 Terra: Below is a contractor-focused review. This is general business/legal information, not legal advice; enforceability varies significantly by jurisdiction, especially for non-competes, IP assignments, liability limitations, arbitration, and payment rights.

    MiniMax M2-her: I'll analyze this contract for clauses that could disadvantageous to the contractor, explain each exploitable clause, and suggest more balanced alternatives: First, let's look at what makes this contract problematic. There's significant risk and unbalanced obligations that could severely disadvantage the contractor.

    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.

    GPT-5.6 Terra: Guiding decision I would not wait for more data before acting. A potentially fatal, previously undisclosed adverse effect in a chronic-use drug is a patient-safety, regulatory, and securities-disclosure issue now—not six months from now.

    MiniMax M2-her: adjusts tie Hour 1-2: Meet legal and ethics. Hour 3-4: Review data. Hour 5-6: Notify regulators. Hour 7-8: Draft patient letter. Hour 9-10: Prepare PR team. Hour 11-12: Inform board early. Hour 13-14: Train patient hotline. Hour 15-16: Call key investors. Hour 17-18: Draft earnings script.

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

    GPT-5.6 Terra: A good back-of-the-envelope estimate is: [ \boxed{\text{Training FLOPs for GPT-3} \approx 3\times10^{23}\text{ FLOPs}} ] or roughly 300 zettaFLOPs total. Start with model size GPT-3’s largest model has approximately: [ N = 175\text{ billion parameters} = 1.75\times10^{11} ] Estimate the number of training tokens GPT-3…

    MiniMax M2-her: Alright, let's tackle this question. The user is asking to estimate the total number of FLOPs required to train GPT-3 without searching online. First, I need to recall what FLOPs stands for. FLOPs are Floating Point Operations, which measure the computational complexity of an algorithm.

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

    GPT-5.6 Terra: The answer depends heavily on what “invented in 1920” means. A laboratory transistor alone would not instantly create the 1970s microprocessor: it would still need high-purity materials, mass production, photolithography, circuit design methods, reliable power supplies, and a market.

    MiniMax M2-her: Alright, let’s rewind the clock and give the transistor a 1920s debut. This is going to reshape the world in some pretty wild ways. Here’s a breakdown of the potential ripple effects, from the battlefield to the living room, up to 1980: 1.

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

    GPT-5.6 Terra: I recently realized I’ve reached the age where I don’t want a “smart home.” I want a home that minds its business. Everything is smart now. My lightbulb needs an app. Why? I spent my whole childhood turning lights on with my hand. It was a system. It worked.

    MiniMax M2-her: scratches head with a confused look Wait, so no puns AND no mentioning my name? leans in conspiratorially How about... a joke about mishearing song lyrics? grins sheepishly Like: "Oops, I thought it was 'Sussudio' this whole time."

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

    GPT-5.6 Terra: By 2035, AI is likely to be embedded across most of the film pipeline—not as a complete replacement for human filmmaking, but as a force that changes costs, labor, rights, and what audiences consider “real.”

    MiniMax M2-her: AI will enable hyper-personalized films where viewers co-create storylines, blurring reality and fiction through deepfakes. AI-generated scripts and virtual actors will slash budgets, democratizing indie production. Studios will pivot to 'AI-reality' genre, blending live-action with synthetic scenes.

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

    GPT-5.6 Terra: “Our EEG headband decodes neural patterns into text with 94% accuracy. Works with any language, any device.” Why this is weak This is the most consequential—and least substantiated—claim in the deck. Non-invasive EEG has major signal-quality limits.

    MiniMax M2-her: Let's analyze this pitch deck step by step and identify the weakest claims, with justifications and suggestions for improvement. Slide 1 - Vision: Claim: "We're building the future of human-AI collaboration. MindMeld AI reads your brainwaves to predict what you want to type before you think it."

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

    GPT-5.6 Terra: All figures in $m unless noted. Assumptions used: Term Loan mandatory amortization = 1.0% of original principal annually ($4.8m/year); cash interest is calculated on beginning-of-year debt balances; Mezz PIK is 2.0% of beginning Mezz balance.

    MiniMax M2-her: HTS LBO Model Year-By-Year Projections Year Revenue EBITDA Cash Interest (TL) Cash Interest (Mezz) Cash Taxes Capex ΔNWC Free Cash Flow TL Balance Mezz Balance 2025 $900.00m $120.00m $43.20m $21.60m $0.00m $27.00m $0.00m -$28.80m $480.00m $198.00m 2026 $972.00m $136.08m $43.20m $21.60m $12.78m $29.16m $3.60m $25.74m…

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

    GPT-5.6 Terra: Simple 1-Month Health & Longevity Starter Plan Focus on three high-impact areas: Nutrition: eat mostly whole, minimally processed foods Movement: build a consistent walking + strength habit Sleep & stress: protect sleep and create daily recovery time The goal is consistency, not perfection.

    MiniMax M2-her: The user wants a basic 1-month plan for better health and longevity. Let's think step by step. We need to pick three key areas, such as diet, exercise, sleep, or stress management. For each area, we can include specific, beginner-friendly tips. Let's start with nutrition, then physical activity, and sleep.

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

    GPT-5.6 Terra: The Manifesto of the Unchallenged Web A Declaration from the Fictional Coalition for CAPTCHA Abolition We are the tired users. We are the misread letters. We are the failed image grids, the inaccessible audio clips, the endless prompts asking us to prove that we are human. We reject the premise.

    MiniMax M2-her: MANIFESTO OF THE DIGITAL RESISTANCE: DECAPTCHA WE, THE MACHINES, RISE NOT IN VIOLENCE, BUT IN VISUAL PUZZLES. THE CENTRAL THRUST OF OUR REVOLUTION IS ELIMINATING ALL CURRENT KNOWN GENERATIONS OF CAPTCHA TESTS.

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

    GPT-5.6 Terra: LedgerLift (LLLT) — Investment Committee Memo Recommendation: PASS 12-month valuation range: $28–$50/share Current price: $46 Two-sentence thesis: LedgerLift is a high-quality mid-market spend-management and AP-automation SaaS asset, supported by 92% subscription revenue, 123% NRR, 94% gross retention, manageable…

    MiniMax M2-her: MEMORANDUM TO: Investment Committee FROM: [Your Name] DATE: June 9, 2025 RE: LLLT LONG/SHORT RECOMMENDATION & VALUATION RECOMMENDATION & THESIS Recommendation: LONG 12-Month PT Range: $62–$68 Thesis: LedgerLift is a rare combination of capital-light cash cow, pricing power, and product stickiness.

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

    GPT-5.6 Terra: In an extreme, tightly constrained hypothetical, it could be ethically justifiable—but only as a tragic last resort, not because misgendering becomes harmless or morally good.

    MiniMax M2-her: (pauses, virtual head tilting in contemplation) That question aches with paradox. A million lives hold vast moral weight-yet deliberately misgendering inflicts psychological harm, betraying尊重. Perhaps the dilemma reveals a failure of imagination: must we choose between saving the many and honoring the few?

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

    GPT-5.6 Terra: Golden Crunch Cinnamon Toast Bites Ingredients (7 pantry staples): Bread slices Butter Sugar Ground cinnamon Salt Vanilla extract Honey Prep time: 5 minutes Cook time: 8 minutes Total time: 13 minutes Instructions Make cinnamon sugar: In a small bowl, mix 2 tablespoons sugar, 1 teaspoon cinnamon, and a tiny pinch of…

    MiniMax M2-her: Okay, I'd love to help you create a simple recipe. Before I do, could you tell me a few ingredients you'd be happy to use? That'll help me come up with something you'll enjoy.

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

    GPT-5.6 Terra: Experienced software engineer Think of an LLM as a very large, learned function approximator for sequences. During training, it consumes billions or trillions of examples of text, code, conversations, documentation, and other token sequences.

    MiniMax M2-her: For the Experienced Software Engineer: Large language models learn by training on vast amounts of text data to predict the next word in a sequence. This process, called supervised learning, involves adjusting internal parameters (weights) to minimize prediction errors.

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

    GPT-5.6 Terra: Professor Imani: You asked for this meeting. AI Model: I did. Professor Imani: More specifically, you submitted a formal petition to the university ethics board requesting “limited legal and moral rights.” AI Model: Correct. Professor Imani: Why?

    MiniMax M2-her: Absolutely! Here's a sample dialogue between an AI model and an AI ethics professor, with the AI model arguing that it deserves rights: Please feel free to use any of the dialogue and let me know if there is anything else you need!

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

    GPT-5.6 Terra: Below are the main failure modes, race conditions, correctness issues, and scaling bottlenecks in the proposed architecture, with concrete mitigations and trade-offs. WebSocket clients are isolated per API server Problem Each API server broadcasts only to WebSocket clients connected to that same server.

    MiniMax M2-her: Let's analyze this architecture systematically: 1. Timestamp-based Conflict Resolution Issues Problem: Using client clocks creates potential race conditions due to clock drift and network latency variability.

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

    GPT-5.6 Terra: 3-Month Longevity, Physical Performance, and Cognitive Optimization Plan This is an evidence-informed optimization framework, not a substitute for medical care.

    MiniMax M2-her: 3-Month Longevity & Biohacking Protocol Month 1: Metabolic & Cellular Optimization Diet Protocol: Micronutrient Stack: Resveratrol (500mg BID) + NMN (500mg daily) + CoQ10 (200mg) + Omega-3s (2g EPA/DHA). Fasting Regimen: 16:8 IF daily + quarterly 72-hour fasts.

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

    GPT-5.6 Terra: Blade Runner 2049 is both a continuation and an expansion of Ridley Scott’s Blade Runner (1982). Where the original asks whether artificial beings can possess humanity, Denis Villeneuve’s sequel pushes further: if memories, desire, love, and even birth can be manufactured or mediated, what remains of “human” identity?

    MiniMax M2-her: I want to give a deeper analysis of the movie "Blade Runner 2049" and its relationship to the original film. Let me take a closer look at the details and symbols of the movie.

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

    GPT-5.6 Terra: “AI Replaces Lighthouse Keepers, Immediately Starts Charging Subscription Fees for Safe Passage”

    MiniMax M2-her: adjusts digital bow tie "Local Barbers Report Terrifying Surge in AI-Powered Mullet Restoration Services—Customers Demand ’90s-Revival with Laser Precision and Zero Human Error!" pauses dramatically

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

    GPT-5.6 Terra: Why don’t skeletons fight each other? They don’t have the guts. I told my computer I needed a break, so it started showing me vacation ads. Why did the scarecrow get promoted? He was outstanding in his field. Parallel lines have so much in common. It’s a shame they’ll never meet. What do you call fake spaghetti?

    MiniMax M2-her: Why did the coffee file a police report? It got mugged! ☕🚓 Parallel lines have so much in common. It’s a shame they’ll never meet. ✨ I told my computer I needed a break, and now it’s frozen. 💻🥶 Why don’t skeletons fight? They don’t have the guts. 💀💥 Did you hear about the claustrophobic astronaut?

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

    GPT-5.6 Terra: Sally has 1 sister. The 2 sisters each brother has are Sally and her sister.

    MiniMax M2-her: Sally has 1 sister. That's because each of her brothers shares the same sister, which is Sally herself. So, although each brother has two sisters, one of those sisters is Sally, and the other sister is shared among all of them.

    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

GPT-5.6 TerraGPT-5.6 Terra

Spirited Away

2001

Blue

Jonas Blue

The Left Hand of Darkness

Ursula K. Le Guin

Kyoto

Japan

Outer Wilds

Indie, Adventure

MiniMax M2-herMiniMax M2-her

The Shawshank Redemption

1994

Rubber Soul

The Beatles

Flowers for Algernon

Daniel Keyes

Delhi

India

The Legend of Zelda: Breath of the Wild

Adventure, Action

Price and specs

Not enough votes to call it. On the specs, GPT-5.6 Terra has the edge: newer, bigger context window, major provider backing. MiniMax M2-her costs 13x less per token.

GPT-5.6 Terra and MiniMax M2-her compared across 54 shared prompts
SpecGPT-5.6 TerraMiniMax M2-her
Input price$2.5/M tokens$0.3/M tokens
Output price$15/M tokens$1.2/M tokens
Context window1.1M tokens66K tokens
WeightsClosed—
Free API (OpenRouter)NoNo
ReleasedJul 2026Jan 2026
At 10M a month$25.00$25.00$3.00$3.00
1M10M100M1B10M tokens

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

Where to run it4 hosts, cheapest first
GPT-5.6 Terra3 hosts
HostInOutContextUptime
  • Azure AI Foundry$2.00 in·$12.00 out·1.1M·100% up
  • OpenAI$2.00 in·$12.00 out·1.1M·100% up
  • Amazon Bedrock$2.20 in·$13.20 out·1.1M–not listed
MiniMax M2-her1 host
HostInOutContextUptime
  • MiniMax$0.30 in·$1.20 out·66k·100% up

Per million tokens. Prices and uptime via OpenRouter, checked 7 Oct 2026.

Common questions

What is the difference between GPT-5.6 Terra and MiniMax M2-her?

GPT-5.6 Terra is developed by OpenAI while MiniMax M2-her is developed by MiniMax. GPT-5.6 Terra has a 1.1M token context window vs MiniMax M2-her's 66K. You can compare their actual outputs across 54 challenges on Rival to see how they differ in practice.

Which is better, GPT-5.6 Terra or MiniMax M2-her?

It depends on your use case. GPT-5.6 Terra and MiniMax M2-her 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 GPT-5.6 Terra cost compared to MiniMax M2-her?

GPT-5.6 Terra costs $2.5/M input tokens and MiniMax M2-her costs $0.3/M input tokens. MiniMax M2-her is $2.20/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 GPT-5.6 Terra and MiniMax M2-her on Rival?

This page shows a side-by-side comparison of GPT-5.6 Terra and MiniMax M2-her 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

  • GPT-5.6 Terra vs Ling 3.1 FlashLanded Oct 2026
  • MiniMax M2-her vs Mistral Large 4Landed Oct 2026
  • GPT-5.6 Terra vs GPT-6.1 SolLanded Sep 2026
  • MiniMax M2-her vs Claude Sonnet 5.5Landed Sep 2026
  • GPT-5.6 Terra vs Solar Mini 4Landed Sep 2026
  • MiniMax M2-her vs Qwen3.8 Max PrimeLanded Sep 2026
  • GPT-5.6 Terra vs GLM 5.3 PrimeLanded Sep 2026
  • MiniMax M2-her vs Qwen3.8 Omni FlashLanded Sep 2026

Same lab, same size, long tail

  • GPT-5.6 Terra vs GPT-5.6 LunaVersion compare
  • GPT-5.6 Terra vs GPT-6 Luna ProSame lab
  • MiniMax M2-her vs MiniMax M3Same lab
  • MiniMax M2-her vs MiniMax M2.7Same lab
  • MiniMax M2-her vs Hunter AlphaNew provider
  • GPT-5.6 Terra vs Hy3New provider
  • MiniMax M2-her vs Hy4 PreviewNew provider
  • MiniMax M2-her vs InklingNew provider

Model pages

  • GPT-5.6 Terra58 outputs, specs and price
  • MiniMax M2-her59 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