Introducing

OspreyHuman-like conversational AI.

A new generation of fast, affordable LLMs built for AI companions, AI assistants and real-time applications.

Try it freeRead the API docs

Free in the Playground today · Public launch 1 October 2026

Not the coding geek.The extrovert of LLMs.

Most language models are built to top coding and exam leaderboards. Osprey is built for people. It talks, listens and understands the everyday — your plans, your preferences, your questions, your day — and handles them like a pro.

That makes it the model for what people actually do with AI: talk to it, ask it things, and have it get things done — in companions, assistants, customer support and real-time voice.

  • Up to 90% human-like
  • ~150 tokens a second
  • From ₹6.9 per 1M tokens
  • Tools on every mode
The Osprey mascot in flight.

Higher intelligence. Faster responses. Lower costs.

Osprey Flash, our flagship — a 552B backbone with 8B prefill and 16B decode — balances intelligence, speed and cost, so an AI that actually feels conversational is practical to build.

~150

tokens / sec

Ultra-low latency

Replies land while you're still reading — fast enough for voice and real-time apps.

48–58

Intelligence Index

High-accuracy reasoning

The highest score of the four models in our benchmark.

90%

human-like

Natural conversations

It sounds like a person — warm, clear and to the point.

Every

mode

Tool calling & agents

Calls your functions and APIs on every Flash and Pro mode.

₹6.9

$0.071 / 1M input

Affordable LLM API

Most affordable in its class. Prepaid in rupees, no subscription.

Every AI can write code.Who takes care of the rest of your life?

The biggest models in the world — Fable, Astra, Kimi — are built for coding and computer use, and they do the rest at 10 to 20 times the cost. Osprey is the LLM built for real life: it connects AI to people, homes, cars, robots and machines.

Made for every day. Talk to it about anything — it gets things done.

Plans, preferences, a bad day, a trip, a late order, a table for four. Osprey keeps the thread, answers like a person and uses your tools to finish the job.

9:41
OspreyDriving · live navigation
LIVE MAPNavigating live

Take me to the airport — and avoid tolls.

car.navigate · live map · no tolls

I'll control the car and navigate you in real time — left at the next signal, then the ring road. 42 min, AC at 22°.

Illustrative conversations. The tools are your own functions, called by Osprey.

Give every machine a mind. The LLM that connects AI to the real world.

Osprey is built to operate machines — IoT devices and home appliances, your car and your bike, AR and VR, humanoid robots, farms, drones and air taxis, factories, even defence. It understands people, talks naturally and calls the right tool, fast enough for the real world.

Humanoid robots

Can you help Dadi with her medicines?

robot.fetch · glass of water

Of course — it's 8 pm, so the blue tablet. I've brought her water too.

Smart home

Movie night.

home.scene · lights 20% · blinds down

Enjoy! Lights dimmed, blinds down, AC at 23°.

Your car

Take me to the airport — avoid tolls.

car.navigate · live

42 minutes, no tolls. I'll guide you and keep the cabin at 22°.

Your bike

Any fuel on the way?

maps.nearby · fuel

2 km ahead on the left — I'll guide you in your helmet.

AR & VR

What's this building?

Built in 1887, one of the city's oldest — want the story while you walk?

Industry

Line 3 is running hot.

plc.set_speed · line 3 · 80%

Spindle at 78 °C — I've slowed it to 80% and alerted maintenance.

Home appliances

Clean the living room before the guests arrive.

vacuum.start · living room · deep clean

On it — a deep clean around the sofa and the rug, done in 25 minutes. I'll dock it before 7.

Agriculture

Which field needs water today?

sensors.soil · field 3 → drone.spray

Field 3 is down to 18% moisture — the drone is watering it now and skipping the rows that are already wet.

Drones & air taxis

Send this parcel to Riya — and get me an air taxi home.

drone.dispatch · parcel → air_taxi.book

The parcel lands on Riya's balcony in 12 minutes, and your air taxi is 4 minutes away.

  • IoT devices
  • Home appliances
  • Wearables
  • Industrial machines
  • Healthcare devices
  • Defence & security

Illustrative scenes. The tools are your own functions, called by Osprey.

Benchmarks. Leading on conversation and intelligence.

Osprey Flash against three models developers compare it with, on MiniCrow's current benchmark. Best quality result in each row in bold.

Swipe the table to compare all four models

Benchmark results for Osprey Flash, GPT-5.4 Luna, DeepSeek V4.1 Flash and Sarvam 105B: human-like conversation score, Intelligence Index (AAI), output speed, and input and output price per million tokens in dollars and rupees.
BenchmarkOsprey Flash Most accurateGPT-5.4 Luna(max)DeepSeek V4.1 FlashSOTASarvam 105B105B sovereign Indian LLM by Sarvam
Quality
Human-like conversation190%78%72%65%
Intelligence Index (AAI)248–58384026
Speed and price
Output speed, tokens/s~150~130~214–266~100
Input price, per 1M tokens3$0.115₹11.05$0.174₹16.72$0.05–$0.444₹4.80–₹42.28$0.35₹33.63
Output price, per 1M tokens3$0.75₹72.07$1.2₹115.32$0.14–$1.324₹13.45–₹126.85$0.88₹84.57
  1. Human-like conversation score: share of conversations rated as human-like in MiniCrow's benchmark.
  2. Intelligence Index (AAI); Osprey Flash is reported as a range across its modes.
  3. Pricing references used in the benchmark, not Osprey's API prices — those are live in Which Osprey. Rupees at ₹96.1 to the dollar.
  4. Off-peak to standard price.

MiniCrow's current benchmark measurements; results vary by workload, prompting, infrastructure and measurement methodology. GPT-5.4 Luna, DeepSeek V4.1 Flash and Sarvam 105B are third-party models named for comparison only; MiniCrow is not affiliated with their makers.

One of the most powerful and affordable LLMs

for making your own JARVIS — an AI companion — even better than GPT-5.4 Luna and DeepSeek V4.1 Flash in our benchmark.

90%

Up to 90% human-like score

Natural conversations. Real understanding. Truly intelligent.

Human-like conversation score
  • Osprey Flash90%
  • GPT-5.4 Luna78%
  • DeepSeek V4.1 Flash72%
  • Sarvam 105B65%
Intelligence Index (AAI)
  • Osprey Flash48–58
  • GPT-5.4 Luna38
  • DeepSeek V4.1 Flash40
  • Sarvam 105B26
Output speed, tokens per second
  • Osprey Flash~150
  • GPT-5.4 Luna~130
  • DeepSeek V4.1 Flash~214–266
  • Sarvam 105B~100
Intelligence against priceHigher is smarter, further left is cheaper.
2030405060$0.00₹0$0.50₹48$1.00₹96$1.50₹144Output price per 1M tokensIntelligence Index (AAI)Osprey Flash — Intelligence Index 48–58, output $0.75 per 1M tokensOsprey FlashGPT-5.4 Luna — Intelligence Index 38, output $1.2 per 1M tokensGPT-5.4 Luna (max)DeepSeek V4.1 Flash — Intelligence Index 40, output $0.14–1.32 per 1M tokensDeepSeek V4.1 Flash SOTASarvam 105B — Intelligence Index 26, output $0.88 per 1M tokensSarvam 105B

Build your own JARVIS.

One of the biggest things Osprey is built for is the AI companion — more than just an AI, a companion that truly understands.

The Osprey mascot rising with its wings raised.
  • Plan my day, like a real assistant.
  • Remember my preferences.
  • Talk naturally, like a friend.

Imagine an AI that can:

  • Remember your context
  • Understand your preferences
  • Have natural conversations
  • Use tools on your behalf
  • Access your personal or professional knowledge
  • Plan tasks
  • Call APIs
  • Execute workflows
  • Respond quickly enough to feel conversational

This opens the door to your own JARVIS-style AI assistant — on an API you can afford to run at scale.

  • AI CompanionsPersonalised. Always there.
  • AI AssistantsPlan. Organise. Get things done.
  • Smart ApplicationsContext-aware. More capable.
  • Developer FriendlySimple APIs. Real-world ready.

Built for real-world applications. Anywhere people talk to machines.

  • Personal AI assistantsCalendars, reminders, errands and answers — in one conversation.
  • AI companionsSomeone to talk to who remembers, understands and cares.
  • Customer support agentsOrder status, refunds and answers, around the clock.
  • Enterprise AI assistantsYour company's knowledge and tools, in plain conversation.
  • Knowledge agentsGrounded answers from your own documents, with RAG.
  • Voice AI applicationsAdd Lark to hear and Pica to speak — a voice that talks back.
  • AI-enabled vehiclesA co-driver you can simply talk to.
  • AR and VR experiencesCharacters that converse in real time.
  • Robotics and humanoid AINatural conversation for machines that move.
  • Tool-using AI agentsPlan, call APIs and run workflows on their own.

Built for developers. Infrastructure, not just a chatbot.

  1. LLM APIs
  2. Streaming
  3. Tool calling
  4. Context
  5. RAG
  6. Agents
  7. Production applications

Switch in two lines

Osprey speaks the OpenAI API: change the base URL and the key, and keep the code you have.

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.minicrow.com/v1",
    api_key="mc_YOUR_KEY",
)

reply = client.chat.completions.create(
    model="osprey-flash",
    messages=[
        {"role": "user", "content": "Plan a 5-day Spiti trip for ₹25,000."},
    ],
)
print(reply.choices[0].message.content)

One platform for the whole assistant

Combine Osprey with the rest of MiniCrow — one key, one bill.

A complete real-time AI application stack, through one platform.

Which Osprey is right for you?

Three models, one API. Live prices per million tokens, prepaid in rupees.

The Osprey mascot, a glowing violet and cyan osprey, banking into a dive.

Coming soon

Osprey Flash Lite

The lightest Osprey.

For volume — classification, routing and quick extraction. Reads images, audio and files.

From ₹9.5 per 1M input tokens

Coming soonDocs
Tool calling
Not available
Modes
Speed · Intelligence
Takes
Text · images · audio · files
Default (intelligence), per 1M
₹9.5 in · ₹35.9 out
$0.099 in · $0.374 out
Best for
High-volume classification · Tagging and routing · Cheap extraction passes
The Osprey mascot rising, wings raised and talons forward.

Most popular

Osprey Flash

552B backbone · 8B prefill · 16B decode

The one to build on.

Fast, capable and affordable, with tools on every mode. Right for assistants, companions and agents.

From ₹6.9 per 1M input tokens

Try itDocs
Tool calling
Every mode
Modes
Speed · Intelligence · Max
Takes
Text · images · files
Default (intelligence), per 1M
₹7.9 in · ₹26.4 out
$0.082 in · $0.275 out
Best for
Chat and assistants · Agent loops with tools · Structured extraction
The Osprey mascot soaring towards you, wings spread wide.

 

Osprey Pro

For work that must be right.

Long agent loops, long-context reasoning and analysis where a mistake costs more than the tokens.

From ₹79.2 per 1M input tokens

Try itDocs
Tool calling
Every mode
Modes
Speed · Intelligence · Max
Takes
Text · images · audio · files
Default (intelligence), per 1M
₹147.8 in · ₹464.6 out
$1.54 in · $4.83 out
Best for
Multi-turn agent loops · Long-context reasoning · Code and analysis that must be right
Every mode and its price
ModelModeForInput · 1MOutput · 1M
Osprey Flash Litecoming soonspeedThinking off. For volume.₹9.5$0.099₹35.9$0.374
Osprey Flash Litecoming soonintelligencedefaultThinking on. The default.₹9.5$0.099₹35.9$0.374
Osprey FlashspeedShort exchanges, low latency.₹6.9$0.071₹19.0$0.198
Osprey FlashintelligencedefaultThe default. Tools, images, most work.₹7.9$0.082₹26.4$0.275
Osprey FlashmaxWhen it has to hold up.₹21.1$0.220₹126.7$1.32
Osprey ProspeedThe widest input of any lane.₹79.2$0.824₹396.0$4.12
Osprey ProintelligencedefaultThe default. Deep reasoning.₹147.8$1.54₹464.6$4.83
Osprey PromaxThe top of the catalogue.₹211.2$2.20₹1,056.0$10.99

Live from our price book; dollars at ₹96.1. Every MiniCrow price

The next generation of AI applications won't be simple chatbots.

They will be AI systems that can see, hear, remember, reason, use tools and communicate naturally. Osprey is our step toward making that infrastructure:

  • Fast.
  • Affordable.
  • Intelligent.
  • Conversational.
  • Built for developers.

Public launch

1 October 2026.

Build › Deploy › Scale

Build without limits. Talk to Osprey free in the Playground today — no account needed.

Try it freeRead the API docs
The Osprey mascot face on, wings raised and tail fanned.

Questions.

What is Osprey?

Osprey is MiniCrow's family of fast, affordable language models built for human-like conversation — the extrovert of LLMs. It is designed for AI companions and personal AI, assistants, context-aware applications, agents and autonomous workflows, tool and function calling, knowledge-based and real-time applications, customer support and enterprise AI.

How is Osprey different from other LLMs?

Most models are tuned to top coding and exam leaderboards. Osprey is tuned for people: everyday conversation, context, preferences and getting things done with tools. In our benchmark Osprey Flash scored up to 90% on human-like conversation and 48–58 on the Intelligence Index, ahead of GPT-5.4 Luna (78%, 38), DeepSeek V4.1 Flash (72%, 40) and Sarvam 105B (65%, 26), at about 150 tokens per second. Results vary by workload, prompting, infrastructure and measurement methodology.

How much does Osprey cost?

Osprey is prepaid in rupees with no subscription, and every response carries its own cost. Osprey Flash costs ₹6.9–₹21.1 per million input tokens ($0.071–$0.220) and ₹19.0–₹126.7 per million output tokens ($0.198–$1.32), by mode. The $0.115 and $0.75 in the benchmark are the pricing references used there.

How big is Osprey Flash?

Osprey Flash: 552B backbone / 8B prefill / 16B decode.

Can I try Osprey for free?

Yes. The MiniCrow Playground lets you talk to Osprey free, without an account. API keys open to everyone on 1 October 2026.

Is Osprey OpenAI-compatible?

Yes. Point the OpenAI SDK at https://api.minicrow.com/v1 with a MiniCrow key and call osprey-flash or osprey-pro (osprey-flash-lite is coming soon). Chat, streaming, tool calling and images use the request shape you already have.

Does Osprey support tool calling?

Yes — Osprey Flash and Osprey Pro call tools in every mode. Osprey is one of the best and most effective models for tool calling on everyday, human-level tasks: bookings, orders, calendars, lookups, reminders and whole workflows, done in conversation.

Can I build my own JARVIS-style assistant with Osprey?

That is one of the biggest use cases Osprey is built for: a companion that remembers your context, understands your preferences, has natural conversations, uses tools on your behalf, plans tasks, calls APIs and executes workflows — quickly enough to feel conversational. Add Lark to hear, Pica to speak and Osprey Live for real-time calls.

When does Osprey launch?

MiniCrow launches Osprey publicly on 1 October 2026. You can talk to it in the Playground today.

Meet the MiniCrow family. One API that talks, hears, speaks and watches.

Osprey thinks; the rest of MiniCrow gives it a voice, ears and eyes. Press play: hear a real call, a recording and its transcript, a voice — or watch Lark-V read a real clip.

Lark

Hears

Speech to text for India — Hindi, Marathi, Tamil and more, measured on real 8 kHz phone calls.

  • speech-to-text API for 21 languages
  • 4 tiers · nano · mini · large · Speaker labels
नमस्ते, मेरा ऑर्डर कब आएगा?

Hindi · 8 kHz phone call → text, in real time

₹20

per hour of audio · $0.208

Playground Live demo

Pica

Speaks

Natural text to speech in Hindi and English, with expressive voices for assistants, IVR and narration.

  • text-to-speech API for Hindi and English
  • 3 tiers · nano · small · large

“Welcome back, Riya! Your order is on its way.”

₹21

per hour of audio · $0.219

Playground Live demo

Lark-V

Watches

Video understanding — a summary, a seekable timeline and the spoken words, from a single clip.

  • video summary and timeline API
  • 3 tiers · nano · mini · large
  • 00:04A car pulls into the driveway
  • 00:12Two people unload boxes
  • 00:31Someone waves at the camera

Per clip

billed on what each clip used

Playground Live demo

Osprey Live

Talks

Real-time multimodal AI that sees, hears, thinks and talks — with tool calling, for AI calling, companions, robots, glasses and cars.

  • real-time multimodal AI model
Live call · 00:42

Caller: “Do you deliver on Sundays?”

Osprey Live: “Yes — until 9 pm. Shall I book you a slot?”

₹65.74

per call-hour, all in · $0.684

Details Live demo

Embeddings & Reranking Finds

Semantic search and RAG — embed your documents, rerank a shortlist, and give Osprey your own knowledge to answer from.

Docs