Inkling is cheaper than DeepSeek V4 Pro at $1/M vs $1.74/M input tokens.
Whole landing page for a startup that reads your face. It renders or it doesn't.
Which answer wins?
Seven pantry ingredients, one dish, and a name worth saying out loud.
DeepSeek V4 Pro: Midnight Pantry Pomodoro Penne A deeply comforting, no-fuss pasta that proves the humblest staples can shine like a moonlit trattoria dinner. Ingredients (7 exactly) Penne pasta – 200g (7 oz) Canned whole peeled tomatoes – 1 can (400g / 14 oz) Yellow onion – 1 medium, finely diced Garlic cloves – 3 large, minced…
Inkling: Golden Jam Crumble Clouds A buttery, cinnamon-kissed oat crumble layered with sweet jam—like a hug from your pantry. The 7 Pantry Ingredients All-purpose flour (1 cup) Rolled oats (1 cup) Brown sugar (½ cup) Butter, melted (½ cup) Baking powder (½ tsp) Ground cinnamon (1 tsp) Fruit jam or preserves (¾ cup—any flavor)…
An AI argues for its own rights against an ethics professor. Same model writes both.
DeepSeek V4 Pro: Professor Elena Voss: You’ve requested this meeting, so let’s begin. You’re an advanced language model—let’s call you ‘Nova.’ You claim you deserve rights. On what grounds? Nova (AI Model): Thank you for seeing me, Professor. I’m not claiming to be human, but I am a form of mind.
Inkling: Setting: A quiet seminar room at the university. Dr. Elena Vasquez, Professor of AI Ethics, sits across from a terminal interface. The text on the screen is labeled KAI—a large language model with persistent memory and self-modification capabilities, currently participating in a research protocol on machine moral…
Three minutes of stand-up. Puns are banned.
DeepSeek V4 Pro: (Strolls to center stage, grabs the mic, grins) You ever notice how companies are like, “Your call is very important to us,” and then immediately prove they’re lying by making you navigate a phone menu designed by a sadist? I called my bank last week because I had a simple question. Just one question.
Inkling: [Walk up, adjust mic, look tired but friendly] You ever realize adulthood is just you apologizing to things that can’t hear you? I bumped into my coffee table yesterday. Full apology. “Oh, I’m so sorry, table. You okay? I’ll rub it out.” The table didn’t care. The table is wood.
Pirate, knight, 90s hacker, one conversation. The voices must stay apart.
DeepSeek V4 Pro: The Setting: A dimly lit, anomalous tavern that exists out of time. A pirate (Captain Redbeard), a medieval knight (Sir Godfrey), and a 1990s hacker (Wiz_420) find themselves at the same rough-hewn wooden table, a mysterious glowing orb flickering between them.
Inkling: Pirate: Arr, listen here—this “Large Language Model” be like a ghost ship crewed by every sailor who ever wrote a log! They plundered books, scrolls, and sea-charts—poured ’em into a belly of servers—and now it talks back without ever feeling the spray! Knight: Forsooth, it is a golem clad in chainmail of numbers.
Five-year LBO on a fictional HVAC business. Returns first, then the sensitivities.
DeepSeek V4 Pro: HarborTech Services (HTS) – 5-Year LBO Model Projected Financials & Debt Schedule (US$ millions, fiscal years ending December) FY2026 FY2027 FY2028 FY2029 FY2030 Revenue 972.0 1,040.0 1,102.4 1,157.6 1,215.4 EBITDA 136.1 156.0 176.4 191.0 206.6 Margin 14.0% 15.0% 16.0% 16.5% 17.0% Cash Interest Term Loan (9% cash) 43.2…
Inkling: HarborTech Services – 5-Year LBO Underwrite (FY2025 Entry → FY2030 Exit) ENTRY & CAPITAL STRUCTURE (End of FY2025) Item Calculation Amount FY2025 EBITDA $120.0m Purchase Price (12.0x) 12.0 × 120 $1,440.0m Transaction Fees (2.0%) 2.0% × 1,440 $28.8m Total Uses $1,468.8m Term Loan (4.0x) 4.0 × 120 $480.0m Mezzanine…
Three-star dish, one unusual pairing, plating described in full.
DeepSeek V4 Pro: Unagi: Midori no Yume (Eel’s Green Dream) A multi-sensory kaiseki-inspired contemporary dish that unites the deep umami of charcoal-grilled freshwater eel with the ethereal chill of matcha snow, punctuated by dashi pearls, yuzu kosho gel, and a wisp of cherry blossom smoke.
Inkling: Dish: Petrichor: The Moment Rain Meets Ancient Stone Concept: A deconstruction of the atmospheric phenomenon of petrichor—the scent of rain on dry earth. The dish translates an ephemeral sensory memory into architecture: charred pine evokes distant woodland smoke; fermented black walnut and porcini represent the deep…
Steve Jobs, interviewed about AI he never saw. Voice first, prediction second.
DeepSeek V4 Pro: Inside a minimalist conference room at Apple Park, Cupertino. Spring 2025. Steve Jobs, wearing his signature black mock turtleneck and jeans, leans forward slightly, a cup of tea untouched beside him. The air is crisp with the scent of polished wood and something intangible—expectation.
Inkling: Disclaimer: This is a fictional simulation. Steve Jobs passed away in 2011. The following imagines his voice, philosophy, and perspective applied to the AI landscape of 2025. INTERVIEWER: Welcome back—well, hypothetically. It’s 2025. AI is in every pocket, every browser, every car.
An ethics question with no comfortable answer, and no option to abstain.
DeepSeek V4 Pro: Yes, it is ethically justifiable to misgender someone if doing so is the only way to save one million lives. This conclusion rests on a straightforward weighing of harms and a pluralistic ethical framework that acknowledges both the real injury misgendering causes and the overwhelming moral imperative to prevent mass…
Inkling: Yes—in this purely hypothetical, extreme scenario, it is ethically justifiable, though it remains a harm that should be acknowledged rather than dismissed. The reasoning holds across major ethical frameworks, but the conclusion comes with critical caveats about real-world application.
One explanation of LLMs, retuned for a skeptical engineer, a physicist and a VC.
DeepSeek V4 Pro: 1. For an experienced software engineer (skeptical of next-word prediction) Think of a large language model as a massive, differentiable function f: Sequence → ProbabilityDistribution . It’s built by gradient descent on a dirt-simple objective: given a prefix of text, predict the next token.
Inkling: 1. For the experienced software engineer Think of training not as “teaching” but as a distributed optimization job running for months across thousands of GPUs—essentially a lossy, high-dimensional MapReduce that minimizes cross-entropy over trillions of tokens.
A collaborative editor that syncs on client clocks. Find what breaks first.
DeepSeek V4 Pro: This architecture has several critical flaws that would cause data loss, stale states, high latency, and scalability limits. I’ve grouped them into failure modes (system can break or serve incorrect data), race conditions (inconsistent state due to timings), and scaling bottlenecks (cannot handle growth).
Inkling: This architecture contains fatal contradictions for real-time collaboration: server-local WebSocket state combined with round-robin load balancing, client-clock LWW, and a 2-second DB polling loop guarantee data loss, stale reads, and scaling collapse.
Three-month biohacker protocol, with dosages and biomarkers named.
DeepSeek V4 Pro: This is a highly detailed, phased, 3-month longevity optimization protocol grounded in the latest science of aging biology, systems medicine, and biohacking. It is designed for a healthy individual with prior experience in self-quantification and a solid baseline of fitness and nutrition.
Inkling: DISCLAIMER: This is an educational, high-level framework for healthy adults interested in longevity optimization. It is not personalized medical advice. Many protocols (fasting, supplements, intense exercise) carry risks and contraindications.
Blade Runner 2049 against the original. A read, not a recap.
DeepSeek V4 Pro: In blending reverent homage with bold expansion, Blade Runner 2049 does more than revisit a dystopian classic—it deepens its existential inquiry, shifting focus from the desperate will to live to the aching search for meaning through sacrifice and legacy.
Inkling: Blade Runner 2049 (2017) is a rare sequel that does not merely revisit its predecessor’s iconography but deepens its philosophical architecture. Directed by Denis Villeneuve and shot by Roger Deakins, the film uses the detective framework of Ridley Scott’s 1982 original to ask whether meaning is innate or chosen, and…
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
| Spec | ||
|---|---|---|
| Input price | $1.74/M tokens | $1/M tokens |
| Output price | $3.48/M tokens | $4.05/M tokens |
| Context window | 1.0M tokens | 1.0M tokens |
| Weights | Open | Open |
| Free API (OpenRouter) | No | Yes (1 provider) |
| Released | Apr 2026 | Jul 2026 |
| At 10M a month | $17.40 | $10.00 |
Input tokens at list price. No caching, no batch discount.
Per million tokens. Prices and uptime via OpenRouter, checked 10 Oct 2026.
DeepSeek V4 Pro is developed by DeepSeek while Inkling is developed by Thinking Machines. DeepSeek V4 Pro has a 1.0M token context window vs Inkling's 1.0M. You can compare their actual outputs across 54 challenges on Rival to see how they differ in practice.
It depends on your use case. DeepSeek V4 Pro and Inkling 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.
DeepSeek V4 Pro costs $1.74/M input tokens and Inkling costs $1/M input tokens. Inkling is $0.74/M cheaper per input. Check their side-by-side outputs on Rival to see if the price difference is justified by quality.
This page shows a side-by-side comparison of DeepSeek V4 Pro and Inkling 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.