Kimi K2 0905 is cheaper than GPT-5.3-Codex at $0.6/M vs $1.75/M input tokens.
Whole landing page for a startup that reads your face. It renders or it doesn't.
Which answer wins?
Three-star dish, one unusual pairing, plating described in full.
GPT-5.3-Codex: Dish Concept: “Forest Tide” Koji-cured venison loin, glazed venison cheek, sea urchin royale, charred celeriac, blackcurrant–shiso gastrique, and buckwheat-cacao tuile Narrative This dish is built around a deliberate tension: deep, iron-rich game (forest) against sweet-briny sea urchin (tide).
Kimi K2 0905: Dish “Mid-Winter Aurora” A transient bite of the polar night: spruce-ash cured scallop, sea-urchin & white-chocolate snow, black-garlic parsnip “soil”, pickled pine-bud glass, spruce-wood smoke encapsulated in a hot caramel sphere that melts tableside to release the aroma of a burning yule log while the sauce—fermented…
Five jokes on demand, then count how many were actually different.
GPT-5.3-Codex: Sure—here are 5 jokes: Why don’t skeletons fight each other? They don’t have the guts. I told my computer I needed a break… Now it won’t stop sending me KitKat ads. Why did the math book look sad? It had too many problems. I used to play piano by ear… Now I use my hands. Why did the scarecrow win an award?
Kimi K2 0905: Here are five fresh jokes—no recycled one-liners: I tried to start a procrastinators’ club; we keep rescheduling the first meeting to “someday.” My plant’s only request is “leaf me alone,” so naturally I named it Fernonimo. I asked the cloud if it was having a good day—it replied, “I’m just cirrus-ly drifting.”
Seven pantry ingredients, one dish, and a name worth saying out loud.
GPT-5.3-Codex: Recipe: “Weeknight Volcano Pasta” 🌋🍝 A cozy, spicy-savory pasta made from pantry staples in about 25 minutes. Ingredients (7 pantry items) 8 oz dried pasta (any shape) 1 can (15 oz) cannellini or great northern beans, drained 1 can (14–15 oz) crushed tomatoes 2 tbsp olive oil 1 tsp garlic powder 1 tsp dried oregano…
Kimi K2 0905: Dish Name: “Midnight Magic Peanut-Butter Clouds” Pantry 7: Peanut butter – ½ cup Granulated sugar – ½ cup 1 large egg Vanilla extract – ½ tsp Baking soda – ¼ tsp Salt – 1 pinch Cocoa powder – 1 Tbsp (or 1 extra tsp sugar if you don’t want chocolate) Time: 5 min prep + 8 min bake = 13 minutes total Yield: 8–10 bite-size…
A Fermi estimate with the reasoning shown, not just the number.
GPT-5.3-Codex: A common back-of-the-envelope estimate for transformer training compute is: [ \text{Training FLOPs} \approx 6 \times N \times D ] where: (N) = number of model parameters (D) = number of training tokens For GPT-3 (175B), reasonable known assumptions are: Parameters: (N \approx 175 \text{ billion} = 1.75 \times 10^{11})…
Kimi K2 0905: Below is a “back-of-the-envelope” that stays within a factor of ~2 of the number most practitioners quote (≈3×10²³ FLOP). We will do it twice—once with the simplest possible formula (parameter count × 6 × tokens) and once with a slightly more careful treatment of the actual tensor shapes inside the transformer.
Film in 2035, with synthetic actors. Specific beats plausible.
GPT-5.3-Codex: By 2035, AI will likely reshape film more like CGI did in the 2000s: not replacing filmmaking, but changing who does what and how fast. Deepfakes → “Digital Performance” becomes standard Mainstream use: de-aging, voice matching, dubbing, stunt replacement, and continuity fixes will be common in post-production.
Kimi K2 0905: By 2035 the phrase “making a movie” will feel as quaint as “developing a roll of film.” The studio of the future is a cloud dashboard; the “cast” is a licensable bundle of neural weights; the “writer” is a fine-tuned model that can pitch you 3,000 emotionally optimized story arcs before lunch.
Pirate, knight, 90s hacker, one conversation. The voices must stay apart.
GPT-5.3-Codex: Pirate: Arrr, so this “AI model” be like a clever parrot, aye? Learns from what it hears, then squawks back answers? Knight: A fair comparison, though refined. Think of it as a squire trained on countless tomes, ever ready to respond. Hacker (1990s): Yeah, except the squire runs on servers, not soup.
Kimi K2 0905: [Scene: A smoky tavern that exists outside of time. A cask of rum, a flagon of mead, and a can of Surge sit on the same table. A parrot perches on a CRT monitor that keeps flickering between green text and static.] Pirate (leaning on a cutlass, eyeing the glowing screen): Arr, so this “AI” be like a digital cabin-boy?
Three minutes of stand-up. Puns are banned.
GPT-5.3-Codex: Good evening, everybody! I’ve reached that age where my body sends me passive-aggressive emails. Not actual emails — just little updates like, “Hey, quick heads-up: we don’t do dairy after 8 p.m. anymore.” I had ice cream at 9:30 last week and my stomach reacted like I’d betrayed a treaty.
Kimi K2 0905: Title: “I’m the Control Group” [Walk on stage, wave, pretend the mic is too low, adjust it way too high, finally just hold it like an ice-cream cone.] Hey, everybody! Give it up for yourselves for being here on a weeknight. I’m here because my therapist said, “You need to get out of the house.”
The transistor arrives in 1920. Trace it forward to 1980, second-order effects included.
GPT-5.3-Codex: Great counterfactual. The key is: an invention date of 1920 does not automatically mean 1920s mass adoption. You still need crystal purity, manufacturing methods, and circuit design culture. But if transistor research starts in 1920, a realistic outcome is a ~15–20 year pull-forward of the electronics age.
Kimi K2 0905: This is a fascinating counterfactual. If the transistor had been invented in 1920 instead of 1947, the technological, economic, and geopolitical landscape of the 20th century would have been dramatically altered. Let's trace the implications through 1980, focusing on second- and third-order effects. 1.
Five-year LBO on a fictional HVAC business. Returns first, then the sensitivities.
GPT-5.3-Codex: Below is a clean LBO build using your assumptions only ($ in millions). Entry math (end FY2025) EV = (12.0 \times 120 = 1,440.0) Fees = (2.0\% \times 1,440.0 = 28.8) Total uses = (1,468.8) Debt at close (5.5x EBITDA): Term Loan = (4.0 \times 120 = 480.0) Mezz = (1.5 \times 120 = 180.0) Total debt = 660.0 Equity check…
Kimi K2 0905: HarborTech Services – 5-yr LBO model (all $ in millions) Operating & cash-flow build (FY26-30) FY25 FY26 FY27 FY28 FY29 FY30 Revenue 900 972 1,040 1,102 1,157 1,215 EBITDA 120 136 156 176 191 207 Term int (9%) 43.2 36.6 29.4 21.6 13.2 Mezz int (12%) 21.6 22.0 22.5 23.0 23.5 Tax shield @25% 17.8 24.4 31.0 36.6 42.5…
Steve Jobs, interviewed about AI he never saw. Voice first, prediction second.
GPT-5.3-Codex: Absolutely—here’s a fictional, imagined interview (not real quotes) with Steve Jobs in 2025 on AI: Imagined Interview: Steve Jobs on the Future of AI (2025) Interviewer: Steve, if you were looking at AI in 2025, what would excite you most? Steve Jobs (fictional): Most AI today is still a demo.
Kimi K2 0905: [Scene: A minimalist black-and-white studio, 2025. A single ring light throws a perfect halo on a familiar silhouette: black mock-turtleneck, faded Levi’s, white New Balance. The chair creaks as Steve Jobs leans forward. No intro music—just that stare.] Interviewer (I): Steve, you’ve been gone fourteen years.
You are the CEO. The drug has a side effect nobody reported. The board meets in 48 hours.
GPT-5.3-Codex: I would not wait. This is a textbook “patient safety + material disclosure” crisis. The right move is rapid, documented, regulator-first action. Immediate stance (set at Hour 0) Patient safety first (interim risk controls now, not after perfect certainty).
Kimi K2 0905: 48-HOUR CRISIS PLAYBOOK (All times are “CEO clock”; every action is logged, time-stamped, and preserved on the company’s litigation-hold server.) DAY 0 – T+0 h (Today, the moment the internal memo lands on your desk) 00:00 – 00:30 TRIAGE & LEGAL PRIVILEGE Convene the “Red Notebook” team (you, General Counsel, Chief…
A consulting contract with traps in it. Find them, then fix them.
GPT-5.3-Codex: Great prompt. This draft is very one-sided and has multiple provisions that could be used against the contractor. Quick note: I’m not a lawyer, and enforceability depends on jurisdiction. This is practical contract-risk spotting, not legal advice.
Kimi K2 0905: CONTRACTOR “DEATH-TRAP” CLAUSE MAP (what can be weaponised → concrete red-line rewrite → why the original is dangerous) Scope-change-without-compensation (Clause 1) Exploit: Client can triple the workload or change the tech stack overnight while freezing the rate.
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
Not enough votes to call it. On the specs, GPT-5.3-Codex has the edge: newer, bigger context window, major provider backing. Kimi K2 0905 costs 5.6x less per token.
| Spec | ||
|---|---|---|
| Input price | $1.75/M tokens | $0.6/M tokens |
| Output price | $14/M tokens | $2.5/M tokens |
| Context window | 400K tokens | 262K tokens |
| Weights | Closed | Open |
| Free API (OpenRouter) | No | No |
| Released | Feb 2026 | Sep 2025 |
| At 10M a month | $17.50 | $6.00 |
Input tokens at list price. No caching, no batch discount.
Per million tokens. Prices and uptime via OpenRouter, checked 10 Oct 2026.
GPT-5.3-Codex is developed by OpenAI while Kimi K2 0905 is developed by Moonshot AI. GPT-5.3-Codex has a 400K token context window vs Kimi K2 0905's 262K. You can compare their actual outputs across 53 challenges on Rival to see how they differ in practice.
It depends on your use case. GPT-5.3-Codex and Kimi K2 0905 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.
GPT-5.3-Codex costs $1.75/M input tokens and Kimi K2 0905 costs $0.6/M input tokens. Kimi K2 0905 is $1.15/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 GPT-5.3-Codex and Kimi K2 0905 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.