The cheapest paid AI model right now is Mistral: Mistral Nemo at $0.03 per million output tokens ($0.03/1M input). In total, 50 models cost under $1 per million output tokens — for context, a million tokens is roughly 750,000 words of AI output, so budget models make high-volume drafting, summarizing, and coding assistance close to free.
Prices below are pulled live from CoreAI's catalog and ranked by output cost, since output tokens dominate the bill in most chat workloads. All of these models are available in the CoreAI app under one subscription — no separate API keys needed.
| Model | Provider | Context | Input $/1M | Output $/1M | Capabilities |
|---|---|---|---|---|---|
| Mistral: Mistral Nemo | Mistral AI | 131K | $0.03 | $0.03 | — |
| Sao10K: Llama 3 8B Lunaris | Sao10k | 8K | $0.04 | $0.05 | — |
| inclusionAI: Ling 3.0 Flash | Inclusionai | 262K | $0.02 | $0.06 | Reasoning |
| Mistral: Ministral 3 8B 2512 (batch) | Mistral AI | 262K | $0.07 | $0.07 | Vision |
| Mistral: Mistral Small 3 | Mistral AI | 33K | $0.05 | $0.08 | — |
| Meta: Llama 3.1 8B Instruct | Meta | 131K | $0.05 | $0.08 | — |
| OpenAI: gpt-oss-20b | OpenAI | 131K | $0.02 | $0.09 | Reasoning |
| Nex AGI: Nex-N2.5-Mini | Nex-agi | 262K | $0.02 | $0.10 | Reasoning, Vision |
| Reka Edge | Rekaai | 16K | $0.10 | $0.10 | Vision |
| Mistral: Ministral 3 3B 2512 | Mistral AI | 131K | $0.10 | $0.10 | Vision |
| Google: Gemma 3 4B | 131K | $0.05 | $0.10 | Vision | |
| MythoMax 13B | Gryphe | 4K | $0.08 | $0.11 | — |
| IBM: Granite 4.0 Micro | Ibm-granite | 131K | $0.02 | $0.11 | — |
| OpenAI: gpt-oss-20b (batch) | OpenAI | 131K | $0.02 | $0.11 | Reasoning |
| Poolside: Laguna XS 2.1 | Poolside | 262K | $0.06 | $0.12 | Reasoning |
| Qwen: Qwen3.7 Flash | Qwen | 1000K | $0.03 | $0.13 | Reasoning, Vision |
| OpenAI: gpt-oss-120b (batch) | OpenAI | 131K | $0.03 | $0.14 | Reasoning |
| Microsoft: Phi 4 | Microsoft | 16K | $0.07 | $0.14 | — |
| Amazon: Nova Micro 1.0 | Amazon | 128K | $0.04 | $0.14 | — |
| Inference.net: Schematron V2 Turbo | Inference-net | 128K | $0.03 | $0.15 | — |
| Inception: Mercury 2.5 | Inception | 260K | $0.04 | $0.15 | Reasoning |
| Qwen: Qwen3.5-9B | Qwen | 256K | $0.10 | $0.15 | Reasoning, Vision |
| Mistral: Ministral 3 8B 2512 | Mistral AI | 262K | $0.15 | $0.15 | Vision |
| Google: Gemma 3 12B | 131K | $0.05 | $0.15 | Vision | |
| Cohere: Command R7B (12-2024) | Cohere | 128K | $0.04 | $0.15 | — |
| Tencent: Hy-MT2-1.8B | Tencent | 8K | $0.04 | $0.18 | — |
| inclusionAI: Ling 3.0 Flash VL | Inclusionai | 131K | $0.06 | $0.18 | Reasoning, Vision |
| Poolside: Laguna S 2.1 | Poolside | 1049K | $0.09 | $0.18 | Reasoning |
| Meta: Llama Guard 4 12B | Meta | 164K | $0.18 | $0.18 | Vision |
| DeepSeek: DeepSeek V4 Flash 0423 | DeepSeek | 1049K | $0.09 | $0.18 | Reasoning |
| Upstage: Solar Mini 4 | Upstage | 524K | $0.05 | $0.20 | Reasoning |
| Meta: Muse Spark 1.3 Contributor | Meta | 1049K | $0.10 | $0.20 | Reasoning, Vision |
| Z.ai: GLM 5.3 Flash (batch) | Z-ai | 1049K | $0.06 | $0.20 | Reasoning, Vision |
| Meta: Muse Spark 1.2 Contributor | Meta | 1049K | $0.10 | $0.20 | Reasoning, Vision |
| NVIDIA: Nemotron 3.5 Lightning | NVIDIA | 262K | $0.07 | $0.20 | Reasoning |
| NVIDIA: Nemotron 3.5 Content Safety | NVIDIA | 131K | $0.20 | $0.20 | Reasoning, Vision |
| Mistral: Ministral 3 14B 2512 | Mistral AI | 262K | $0.20 | $0.20 | Vision |
| OpenAI: GPT-5 Nano (batch) | OpenAI | 400K | $0.02 | $0.20 | Reasoning, Vision |
| ByteDance: UI-TARS 7B | Bytedance | 128K | $0.10 | $0.20 | Vision |
| Google: Gemini 2.5 Flash Lite (batch) | 1049K | $0.05 | $0.20 | Reasoning, Vision | |
| OpenAI: GPT-4.1 Nano (batch) | OpenAI | 1048K | $0.05 | $0.20 | Vision |
| Reka Flash 3 | Rekaai | 66K | $0.10 | $0.20 | Reasoning |
| Qwen: Qwen2.5 7B Instruct | Qwen | 33K | $0.10 | $0.20 | — |
| Meta: Llama 3.2 1B Instruct | Meta | 60K | $0.03 | $0.20 | — |
| inclusionAI: Ling 3.1 Flash | Inclusionai | 262K | $0.07 | $0.22 | Reasoning |
| Inference.net: Schematron V2 Small | Inference-net | 128K | $0.05 | $0.23 | — |
| NVIDIA: Nemotron 3 Nano 30B A3B | NVIDIA | 262K | $0.06 | $0.24 | Reasoning |
| Qwen: Qwen3 14B | Qwen | 41K | $0.12 | $0.24 | Reasoning |
| Amazon: Nova Lite 1.0 | Amazon | 300K | $0.06 | $0.24 | Vision |
| Anthropic: Claude Haiku 5.5 (batch) | Anthropic | 1000K | $0.05 | $0.25 | Reasoning, Vision |
Mistral: Mistral Nemo is currently the cheapest paid model at $0.03 per million output tokens. Free-tier models cost nothing at all — see CoreAI's free AI models page for those.
Models are priced per token — roughly ¾ of a word. Providers charge separately for input tokens (your prompt plus conversation history) and output tokens (the model's reply). Prices on this page are normalized to dollars per 1 million tokens so models can be compared directly.
Modern budget models are surprisingly capable — many are distilled or smaller versions of flagship models and handle everyday chat, summarization, and routine coding well. Flagships still lead on complex reasoning, so a common strategy is using a cheap model by default and switching up for hard problems.
No. CoreAI plans include a monthly usage allowance — you pick any model, and usage draws from your plan's budget. There are no separate API bills or per-token invoices to manage.
Chat with GPT-5, Claude, Gemini and hundreds more — all in one app.