Every Model, One API Key
Call any model below through the OpenAI-compatible MixerLead API at https://api.mixerlead.com/v1. Credits used = (prompt + completion tokens) × the model's credit multiplier.
| Model | Model ID | Type | Context | Credit × |
|---|---|---|---|---|
Meta · Fast, low-cost workhorse for chat, summaries, extraction and agents. |
mixerlead-ai/meta/llama-3.1-8b-instruct-fast | chat | 131K | 1× |
Meta · Meta's mixture-of-experts Llama 4 with a 131K context window, tool calling and vision. |
mixerlead-ai/meta/llama-4-scout-17b-16e-instruct | chat | 131K | 0.75× |
Meta · Meta's flagship 70B instruct model. Strong general reasoning, coding and tool calling. |
mixerlead-ai/meta/llama-3.3-70b-instruct-fp8-fast | reasoning | 24K | 3× |
Meta · Meta's instruction-tuned multimodal model optimized for visual recognition, OCR, image reasoning, and document analysis. |
mixerlead-ai/meta/llama-3.2-11b-vision-instruct | reasoning | 128K | 1.25× |
Mistral AI · Mistral's 24B instruct model with 128K context, vision and tool calling. |
mixerlead-ai/mistralai/mistral-small-3.1-24b-instruct | chat | 128K | 0.5× |
OpenAI · OpenAI's open-weight reasoning model for production agents and complex tasks. |
mixerlead-ai/openai/gpt-oss-120b | reasoning | 128K | 0.75× |
| NVIDIA Nemotron 3 120B MoE NVIDIA · NVIDIA's hybrid mixture-of-experts model with leading accuracy for multi-agent enterprise applications and complex reasoning. |
mixerlead-ai/nvidia/nemotron-3-120b-a12b | Legacy | 33K | 0.75× |
OpenAI · Lower-latency OpenAI open-weight reasoning model. |
mixerlead-ai/openai/gpt-oss-20b | reasoning | 128K | 0.25× |
DeepSeek · Reasoning model distilled from DeepSeek R1. Shows its thinking; best for math and logic. |
mixerlead-ai/deepseek-r1-distill-qwen-32b | Legacy | 80K | 0.75× |
Qwen · Alibaba's frontier reasoning and thinking model capable of competitive math, logic puzzles and step-by-step problem-solving. |
mixerlead-ai/qwen/qwq-32b | Legacy | 33K | 0.75× |
Qwen · Alibaba's 27B instruction-tuned model for multimodal reasoning, high-efficiency generation and agentic workloads. |
mixerlead-ai/qwen/qwen3.8-27b | chat | 33K | 2.5× |
Qwen · Efficient Qwen3 mixture-of-experts model with reasoning and tool calling. |
mixerlead-ai/qwen/qwen3-30b-a3b-fp8 | Legacy | 33K | 0.25× |
Qwen · Code-specialised Qwen model for generation, review and refactoring. |
mixerlead-ai/qwen/qwen2.5-coder-32b-instruct | coding | 33K | 1× |
Zhipu AI · Fast multilingual model (100+ languages) with multi-turn tool calling. |
mixerlead-ai/zai-org/glm-4.7-flash | chat | 131K | 0.5× |
| SEA-LION v4 27B Instruct AI Singapore · Southeast Asian language specialist pre-trained and instruct-tuned for regional linguistic nuances and multilingual tasks. |
mixerlead-ai/aisingapore/gemma-sea-lion-v4-27b-it | chat | 8K | 0.75× |
Meta · Small, quick Meta model for simple replies, tagging and sorting at low cost. |
mixerlead-ai/meta/llama-3.2-3b-instruct | Legacy | 80K | 0.25× |
IBM · IBM's tiny, very low-cost model for RAG, classification and tool calling. |
mixerlead-ai/ibm-granite/granite-4.0-h-micro | fast | 131K | 0.25× |
Meta · Meta's smallest model: very fast and cheap for yes/no answers, labels and routing messages. |
mixerlead-ai/meta/llama-3.2-1b-instruct | Legacy | 60K | 0.25× |
Meta · Meta's pre-trained content safety classifier for input prompt validation and LLM response moderation. |
mixerlead-ai/meta/llama-guard-3-8b | fast | 8K | 0.25× |
BAAI · English embeddings (1024 numbers per text) for search, FAQ matching and similarity. |
mixerlead-ai/baai/bge-large-en-v1.5 | embedding | 1K | 0.2× |
BAAI · High-speed English embedding model transforming text into 768-dimensional vectors for semantic search and retrieval. |
mixerlead-ai/baai/bge-base-en-v1.5 | embedding | 1K | 0.08× |
BAAI · Ultra-compact English embedding model producing 384-dimensional vectors for lowest-latency semantic search. |
mixerlead-ai/baai/bge-small-en-v1.5 | embedding | 1K | 0.05× |
BAAI · Multilingual embeddings, 1024 dimensions, 60K input tokens. |
mixerlead-ai/baai/bge-m3 | embedding | 60K | 0.1× |
Google · Google's state-of-the-art open embedding model trained on 100+ languages, generating 768-dimensional vectors for semantic search. |
mixerlead-ai/google/embeddinggemma-300m | Legacy | 2K | 0.1× |
Qwen · Compact multilingual embeddings, 1024 dimensions, 8K input tokens. |
mixerlead-ai/qwen/qwen3-embedding-0.6b | embedding | 8K | 0.1× |
| PLaMo Embedding 1B (Japanese) Preferred Networks · Premier Japanese and cross-lingual text embedding model producing high-precision 2048-dimensional vectors. |
mixerlead-ai/pfnet/plamo-embedding-1b | embedding | 8K | 0.15× |
Mistral AI · Deprecated upstream. Use mixerlead-ai/mistralai/mistral-small-3.1-24b-instruct instead. |
mixerlead-ai/mistral/mistral-7b-instruct | Legacy | 33K | 1× |