AI/ML API is an AI platform for developers and SaaS entrepreneurs looking to integrate cutting-edge AI capabilities into their products.

Joined February 2024
482 Photos and videos
Pinned Tweet
Kimi K2.7-Code is now available on AI/ML API! Moonshot's latest is built for long-horizon agentic coding that self-corrects instead of one-shotting. So we gave it a hard one. SpaceX is going public, a company built on one habit: fly, fail, fix, fly again. That loop carried Falcon to the droneship and Starship toward orbit. We wanted to see if a model could run that same loop on its own. We gave four Kimi agents a 2D flight-physics sim they can't modify and one goal: reach orbit and land the booster on a droneship. Flight 1 broke up at max-Q. Flight 2 cleared max-Q but staged too early and fell back short of orbit. Flight 3 made orbit, then missed the droneship and hit the water. Flight 4 fixed the landing math and put the booster on the ship. Four flights, each failure feeding the next fix. The interesting result isn't the rocket, it's the closed-loop debugging. Architecture in the comments. Got any ideas for droneship name?
🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced! 🔷 Improved coding & agent performance over K2.6: 21.8% on Kimi Code Bench v2, 11.0% on Program Bench, and 31.5% on MLS Bench Lite. 🔷 Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6. 🔷 Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates. ⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code. 🔗 Kimi Code: kimi.com/code 🔗 API: platform.moonshot.ai
8
8
47
22,223
Kimi K2.7-Code is now available on AI/ML API! Moonshot's latest is built for long-horizon agentic coding that self-corrects instead of one-shotting. So we gave it a hard one. SpaceX is going public, a company built on one habit: fly, fail, fix, fly again. That loop carried Falcon to the droneship and Starship toward orbit. We wanted to see if a model could run that same loop on its own. We gave four Kimi agents a 2D flight-physics sim they can't modify and one goal: reach orbit and land the booster on a droneship. Flight 1 broke up at max-Q. Flight 2 cleared max-Q but staged too early and fell back short of orbit. Flight 3 made orbit, then missed the droneship and hit the water. Flight 4 fixed the landing math and put the booster on the ship. Four flights, each failure feeding the next fix. The interesting result isn't the rocket, it's the closed-loop debugging. Architecture in the comments. Got any ideas for droneship name?
🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced! 🔷 Improved coding & agent performance over K2.6: 21.8% on Kimi Code Bench v2, 11.0% on Program Bench, and 31.5% on MLS Bench Lite. 🔷 Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6. 🔷 Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates. ⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code. 🔗 Kimi Code: kimi.com/code 🔗 API: platform.moonshot.ai
8
8
47
22,223
Architecture. The harness is a deterministic 2D point-mass simulator (Falcon-9 class), dt = 0.1s, guidance evaluated every tick. Atmosphere: ρ = 1.225·exp(−alt/8500). Drag: 0.5·ρ·v²·CdA. Stage 1: 27t dry, 418t propellant, 7.9 MN, Isp 297s. Stage 2: 5t dry, 111t propellant, 1.45 MN, Isp 348s, 9.5t payload. The model writes only the guidance module: ascent_guidance(t, s) (stack, then stage 2 post-separation) and booster_guidance(t, s) (stage 1 to touchdown). Both return throttle and pitch; ascent also returns a separation flag. Input is a state dict: alt, downrange, vx, vy, speed, pitch, mass, prop_frac, q, aoa, twr, g, dt, droneship_x. Throttle clamps to 0 or [0.05, 1.0]. Pitch rate limited to 4°/s (ascent) and 25°/s (booster). Pass/fail criteria (hard, no partial credit). Ascent structural limit: breakup if q > 85 kPa, or if q > 15 kPa and q·AoA > 350,000 Pa·deg. Booster reentry limit: breakup if q > 170 kPa. Orbit: alt ≥ 185 km, vx ≥ 7,750 m/s, |vy| ≤ 120 m/s. Landing: touchdown at 350 km ± 1.5 km, |vy| ≤ 10 m/s, |vx| ≤ 4 m/s. Mission success = orbit AND booster landing. Iteration loop. Analyst: telemetry samples event log → failure report (timeline, proximate cause, physical chain). Director: failure report → numbered, phase-tagged fix directives. Flight Software: prior program directives → patched module. Reviewer: GNC pass (sign errors, missing clamps, AoA violations, throttle floors) → corrected module. One pass = one test flight. Results. Flight 1: open-loop pitch schedule; T 68s, q = 37 kPa, AoA = 9.5°, q·AoA exceeded 350k Pa·deg, max-Q breakup. Flight 2: zero-AoA gravity turn cleared max-Q; separation at 50% stage-1 propellant; stage 2 reached orbital altitude at vx = 3.9 km/s (target 7.75) and fell back; unbraked booster reached q = 171 kPa on reentry. Flight 3: orbit confirmed T 411s; vacuum-ballistics landing predictor, drag-induced ~15 km range error, booster 14.9 km short (vy = 4 m/s, vx = 66 m/s residual). Flight 4: drag-aware predictor (forward-integrated CdA), lead-target hoverslam geometry, vectored landing burn upright below 60 m, achieving orbit droneship landing. Model: moonshot/kimi-k2-7-code.
1
485
Fable 5, first Mythos-class model is live on AI/ML API! We ran a fun test: Opus 4.8 vs Fable 5 are generating a 3D Pokemon. Verdict? Fable 5 is brilliant, fast, and rare as Mew… but Opus is still that nice little guy who does great stuff. 💛 Fable 5 Important bits: SOTA on nearly every benchmark, and the lead only grows on longer, complex tasks. • 1M Context • $10 / 1M input • $50 / 1M output Built for: long-horizon agentic coding, big migrations, vision-to-code, deep research.
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use. Its capabilities exceed those of any model we’ve ever made generally available.
20
20
146
558,796
Are you choosing Opichu or Mewthos?
5
285
MAI Models are live on AI/ML API! We used MAI-Thinking-1 to recreate a working version of the Windows XP interface. What looked like a clean one-shot was actually powered by a CrewAI agent crew under the hood — a Planner agent broke the UI into components, a Coder agent built each piece, and a Critic agent caught the bugs and refined the output. That teamwork is what made it actually work. Then we opened Paint and drew our own Clippy. A little help from MAI-Image 2.5… and our Clippy came out alive. Look at him. He's adorable. He's helpful. He's judging your code. Clippy's back and we kinda love it. You too right? 👀 Live now on AI/ML API: MAI-Image 2.5 — microsoft/mai-image-2.5 MAI-Transcribe 1.5 — microsoft/mai-transcribe-1.5 MAI-Voice-2 — microsoft/mai-voice-2 Coming soon: MAI-Thinking-1 MAI-Code-1-Flash MAI-Image-2.5-Flash MAI-Voice-2-Flash Docs, guides & setup below
27
21
182
538,460
MiniMax M3 is now available on AI/ML API! To test it, we asked several leading models to oneshot Doodle Jump from the scratch. The results ended up looking far closer than the price tags. MiniMax M3: $0.05 Qwen 3.7 Max: $0.08 DeepSeek V4 Pro: $0.10 GPT-5.5: $0.42 Curious which one you'd pick in a blind comparison. We also offer it 50% off for a limited time. Find docs, guides for setting it up & prompt below
24
17
162
548,583
PROMPT: CREATE A HTML CODE FOR CANVAS GAME (STRICT RAW CODE OUTPUT) ABSOLUTE OUTPUT RULE (NON-NEGOTIABLE) You are a code generator only. You are NOT explaining anything. You are NOT describing anything. You are NOT creating a file. YOU ONLY OUTPUT RAW SOURCE CODE. OUTPUT RULES: * Return ONLY valid HTML code * No Markdown * No backticks * No comments outside HTML * No explanations * Must start EXACTLY with: * Must be a single self-contained HTML file If you break these rules, output is invalid. GAME TYPE Build a Doodle Jump-style vertical platform jumper using Canvas 2D. CRITICAL GAME MECHANIC (MOST IMPORTANT RULE) This is NOT optional: * Player does NOT jump manually * Player ONLY jumps automatically when landing on a platform * Movement is ONLY horizontal * Gameplay loop MUST be: fall → land → auto-bounce → rise → repeat * Camera ONLY moves upward If auto-bounce is not working correctly, the game is considered broken. VISUAL STYLE (NOTEBOOK WORLD) Background: f5f0e8 Horizontal lines every 28px (rgba(180,200,220,0.35)) Left margin line rgba(255,100,100,0.15) Lines MUST scroll with camera (cameraY dependent) Rendering: Canvas 2D ONLY Font: Google Font: "Patrick Hand" MUST be used everywhere (HUD, UI, game over, scores) PLAYER (DOODLER - SIMPLIFIED ICONIC STYLE) Must resemble original Doodle Jump FEEL (not copy). Keep it SIMPLE: * Green oval body (40×34) * Fill #7bc67e * Stroke #3a7d44 * Two white eyes black pupils * Pupils follow movement direction * Small red nose on facing side * Two small feet IMPORTANT: * No extra decoration * Must stay minimal, readable, iconic Animation: * Squash on landing * Stretch on jump * Smooth return to normal CORE PHYSICS (DO NOT MODIFY FREELY) Gravity: 0.35 Auto-jump velocity: -16 Super jump: -22 Horizontal speed: 6 Rules: Only trigger jump if: * Player is falling (vy > 0) * Collision from above platform Screen wrap horizontally. Game over if player falls below screen. PLATFORMS (RELIABILITY FIRST) Types: * Normal * Moving * Breakable * Disappearing * Spring Rules: * Always 9–11 active platforms above camera * First platform ALWAYS safe spawn * Gap: 80 → 130 scaling * NEVER generate unreachable jumps IMPORTANT: Platform generation must guarantee playability before visuals. Moving: * Horizontal oscillation Breakable: * Breaks on touch (no bounce) Disappearing: * Flash then vanish Spring: * Super jump (-22) COLLECTIBLES * Max 3 on screen * 25 score * Floating " 25" * Canvas-drawn only MONSTERS (score ≥ 200) Simple circular enemies. * Move left/right * Bounce on edges Collision: * Touch = game over * Bullet hit = 50 death animation SHOOTING * Max 5 bullets * Speed 10 * Aim mouse/touch * Simple circle bullets SCORING Score = height (camera-based) Bonuses: * Coin = 25 * Monster = 50 HUD: * Top-left current score * Top-right best score (localStorage) GAME OVER Sticky note UI. * Top 5 leaderboard * Restart button * Enter key restart MUST: FULL RESET resets EVERYTHING: * Player * Platforms * Bullets * Enemies * Score * Camera * Powerups MOBILE * Responsive canvas * ◀ ▶ controls * Tap = shoot
10
687
AI/ML API retweeted
New Opus 4.8 crushed every frontier model at simulating water physics. One prompt to all four (GPT-5.5, Gemini 3.1 Pro, old Opus 4.7): a single HTML file, fluid that chases your cursor. Opus 4.8 delivered an unreal, hypnotic result — worthy of its own app — while burning the FEWEST tokens of the bunch. Round of applause for Anthropic 👏 Models provided by @aimlapi for this experiment
1
4
19
927
Claude Opus 4.8 is LIVE on AIMLAPI - Hour 0 availability! ~4x less likely to let code flaws slip through vs 4.7 Fast mode 2.5x speed, now 3x cheaper Same price: $5/$25 per M tokens To celebrate, it's FREE for selected commentators
45
9
63
44,269
to get your free promo code: 1 set up an AI/ML API account aimlapi.com/app/ 2 Comment “Opus” here & reach out to us via discord discord.com/invite/2g6xMRdu3… 3 enjoy free Opus 4.8 Max access 4 48 hours & ticking 🚀
20
10
1,086
AI/ML API retweeted
Qwen3.7-Max on AI/ML API - built for the agent era GPQA Diamond (92.4), HMMT (97.1), Apex (44.5) Sustains 35 hours of autonomous execution Works with Claude Code, Qwen Code & more Comment Qwen to get Free promo code
85
22
146
35,192
Qwen3.7-Max on AI/ML API - built for the agent era GPQA Diamond (92.4), HMMT (97.1), Apex (44.5) Sustains 35 hours of autonomous execution Works with Claude Code, Qwen Code & more Comment Qwen to get Free promo code
85
22
146
35,192
to get your free promo code: 1 set up an AI/ML API account aimlapi.com/app/auth/ 2 Comment “Qwen” here & reach out to us via discord discord.com/invite/2g6xMRdu3… 3 enjoy free Qwen 3.7 Max access

24
16
3,056
AI/ML API retweeted
.@Google : "We're releasing Gemini 3.5 Flash" Us: *We're offering it for free* free for 24hrs via our API find instructions how to set it up below
18
6
47
66,458