AI Providers Landscape

OpenAI remains the broadest stack for agentic coding and multimodal app building. The current line spans GPT-5.5 / GPT-5.4 frontiers, GPT-5 mini-nano for cheaper throughput, Codex for repo work, realtime/audio APIs, and open-weight gpt-oss.

• GPT-5.5 / GPT-5.5 pro - flagship reasoning and coding
• GPT-5.4 / GPT-5.4 pro - affordable frontier workhorse
• GPT-5.4 mini / GPT-5.4 nano - low-cost, high-volume tiers
• GPT-5 mini / GPT-5 nano - cheaper low-latency GPT tiers
• GPT-5.3-Codex - agentic coding and repo operations
• gpt-realtime-2 / gpt-realtime-1.5 / gpt-realtime-mini / gpt-realtime-translate / gpt-realtime-whisper - realtime voice and transcription
• gpt-audio-1.5 / gpt-audio / gpt-audio-mini - audio-in, audio-out workflows
• gpt-oss-120b / gpt-oss-20b - open-weight reasoning line

• Agentic coding and repo operations
• Realtime multimodal and voice products
• High-iteration prototypes with tools, speech, and open weights

Anthropic's current Claude line is concentrated around Opus 4.7, Sonnet 4.6, and Haiku 4.5. Opus is the long-horizon specialist, Sonnet is the balanced workhorse, and Haiku is the low-latency tier.

• Claude Opus 4.7 / Opus 4.6 / Opus 4.1 - top-end reasoning and coding
• Claude Sonnet 4.6 / 4.5 / 4 - balanced high-performance workhorse
• Claude Haiku 4.5 / 3.5 - fast lightweight tiers

• Complex coding and refactoring
• Long-running agent tasks
• Writing, analysis, and planning

Google's flagship model family is Gemini. The current public stack is centered on Gemini 3 Pro Preview, with Gemini 2.5 Pro / Flash / Flash-Lite beneath it and Gemma as the open-weight line.

• Gemini 3 Pro Preview / Gemini 3 Pro Image Preview - flagship multimodal models
• Gemini 2.5 Pro - advanced reasoning and long-context work
• Gemini 2.5 Flash / Gemini 2.5 Flash-Lite - lower-cost, high-throughput tiers
• Gemma 3 / Gemma 3n - open-weight Google family

• Multimodal reasoning and document synthesis
• Long-context workflows
• Search, Workspace, and device-first apps

xAI is a Grok-centric stack with chat, coding, voice, and media APIs. Treat it as a product platform: Grok 4.3 is the current general model, Grok Build 0.1 is the coding line, and Voice / Imagine cover audio and media.

• Grok 4.3 - current flagship text model
• Grok Build 0.1 - coding and agentic workflows
• Voice API - realtime conversation, TTS, and STT
• Imagine API - image and video generation

• Conversational products with search
• Agentic coding and tool use
• Voice and media generation

Europe's strongest independent model lab. Mistral is the best choice when you want efficient multilingual models, open weights, enterprise deployment, and serious code / document / speech tooling.

• Mistral Large 3 - flagship open-weight multimodal model
• Mistral Medium 3.5 - frontier-class multimodal workhorse
• Mistral Small 4 - compact general-purpose model
• Ministral 3 - edge-friendly 3B / 8B / 14B family
• Magistral Medium 1.2 / Small 1.2 - reasoning-focused family
• Devstral 2 - frontier code agent model
• Codestral - low-latency code generation
• OCR 3 Premier / Voxtral Mini Transcribe 2 / Voxtral Mini Transcribe Realtime / Voxtral TTS - document and audio stack

• Multilingual European workloads
• Enterprise deployment and customization
• Efficient coding and agent workflows

DeepSeek remains the cost/performance pressure valve in the market. The current API line is V4-Pro / V4-Flash, with V3.2 and V3.1-Terminus still important for migration and compatibility.

• DeepSeek-V4-Pro / V4-Flash - current API line
• DeepSeek-V3.2 / V3.2-Speciale - reasoning-first successor family
• DeepSeek-V3.1 / V3.1-Terminus - compatibility and migration line
• DeepSeek-R1 / R1-0528 - legacy reasoning reference line

• Reasoning-heavy workloads
• Code generation and analysis
• High-value deployments with tight cost control

One of the broadest stacks anywhere. Qwen spans proprietary flagships, open-weight multimodal models, coding, translation, audio, image, and retrieval families.

• Qwen3.5-Plus / Qwen3.5-Flash - flagship multimodal general models
• Qwen3-Max - current text flagship
• Qwen3-VL-Plus / Qwen3-VL-Flash - vision-language families
• Qwen3-Omni-Flash - text, image, audio, and video in one model
• Qwen3-ASR-Flash / Qwen3-TTS-Flash - speech recognition and speech synthesis
• Qwen3.5-LiveTranslate-Flash - simultaneous interpretation
• Qwen3-Coder - agentic coding
• Qwen-MT, Qwen-Image-Plus, Qwen-Image-Edit - translation and image stack

• Chinese-English and multilingual workflows
• Open-weight experimentation
• Vision, translation, speech, and coding tools

The dominant open-weight ecosystem outside China. Meta's current headline models are Llama 4 Scout and Maverick, with Llama Guard 4 as the safety line.

• Llama 4 Scout - natively multimodal, long-context
• Llama 4 Maverick - multimodal general-purpose model
• Llama Guard 4 - safety and policy filtering
• Llama 3.2 / Llama 3.1 - legacy broad-deployment baseline

• Self-hosting and fine-tuning
• Open-weight multimodal apps
• Safety filtering and local control

Enterprise-first provider focused on retrieval, grounding, multilingual control, and private deployment. Command A+ is the newest open-source workhorse, while Command A Reasoning, Rerank, Embed, and Transcribe stay central to the stack.

• Command A+ - newest open-source enterprise workhorse
• Command A / Command A Reasoning - flagship enterprise models
• Command A Vision / Command A Translate - multimodal and translation lines
• Command R7B / Command R+ / Command R - retrieval and grounding family
• Aya Expanse / Aya Vision - multilingual and multimodal lines
• Embed / Rerank / Transcribe - retrieval and audio stack

• RAG and enterprise search
• Grounded assistants
• Private and sovereign deployments

A production platform rather than a single model lab. Bedrock is the control plane; Nova is Amazon's own model line for multimodal, reasoning, and speech workloads.

• Amazon Nova 2 Omni / Nova 2 Pro / Nova 2 Lite
• Nova 2 Sonic - speech and conversational voice
• Nova Multimodal Embeddings - semantic retrieval
• Third-party model access via Bedrock
• Agent and guardrail tooling around model use

• AWS-native production apps
• Enterprise governance and routing
• Multi-provider deployments

Microsoft's small-model family is optimized for strong performance per parameter and edge use cases. The current work centers on compact reasoning, multimodal SLMs, and vision reasoning.

• Phi-4-reasoning-vision-15B - compact multimodal reasoning
• Phi-4-reasoning / Phi-4-reasoning-plus - compact reasoning
• Phi-4 - small-model flagship
• Phi-4-mini - compact variant
• Phi-4-multimodal - vision, audio, text
• Phi-3.5 - still widely used in light-footprint setups

• Small-footprint deployments
• STEM-heavy tasks
• Edge and local inference

Baidu's China enterprise stack is still one of the clearest in the market, with a strong emphasis on search, multimodal understanding, and document-heavy workflows.

• ERNIE 5.0 / ERNIE 5.0 Thinking / ERNIE 5.0 Preview
• ERNIE X1.1 / X1.1 Preview
• ERNIE X1 Turbo / X1 Turbo Preview
• ERNIE 4.5 Turbo / ERNIE 4.5 Turbo VL
• PaddleOCR-VL / PP-StructureV3 - document stack

• Chinese enterprise deployments
• Search-augmented assistants
• Document and multimodal knowledge work

One of China's strongest general-purpose model lines, with a clear push into long-horizon coding, multimodal reasoning, and image generation.

• GLM-5.1 / GLM-5.1-HighSpeed - flagship agentic coding line
• GLM-4.7 / GLM-4.7-FlashX - high-intelligence general line
• GLM-5V-Turbo / GLM-4.6V-Flash - multimodal coding and vision
• GLM-4.6 / GLM-4.5 - still used in broader deployments
• CogView-4 - image generation

• Agentic coding and long tasks
• Chinese-language workflows
• Multimodal app stacks

A fast-moving China-side assistant platform with a strong long-context and coding angle. Kimi is increasingly relevant for agentic workflows.

• Kimi K2.6 - latest flagship
• Kimi K2.5 - multimodal general model
• Kimi K2-thinking / Kimi K2-thinking-turbo - reasoning models
• moonshot-v1 vision previews - older long-context line still in circulation

• Long-context assistants
• Chinese first-party product experiences
• Agent and code-heavy workflows

MiniMax is broader than many outsiders realize: text, speech, video, and music are all active product lines.

• MiniMax M2.1 / M2.1-lightning / M2 - text and agent models
• MiniMax M2-her - role-play and long-dialogue model
• MiniMax Speech 2.6 / 2.6 HD - voice agents
• MiniMax Hailuo 2.3 / 2.3 Fast - video
• MiniMax Music 2.6 / 2.5+ - music generation

• Voice and media products
• Agentic coding
• Multimodal consumer apps

Tencent's Hunyuan stack is moving through a major refresh. The current public signal is Hy3 preview for language work, Hunyuan-role-latest for role-play, and HY-3D-3.1 for 3D.

• Hy3 preview - latest flagship language / reasoning model
• Hunyuan-role-latest - role-play and conversational character model
• HY-MT2-Pro - translation-focused model
• HY-3D-3.1 / HY-3D-3.0 - 3D generation
• HunyuanVideo / HunyuanVideo-Avatar - video generation

• 3D, simulation, and interactive content
• Creative tooling
• Media-heavy AI experiences

Runtime and ecosystem layer, not a model lab. It is the simplest way to run open models locally, switch families, and keep work private.

• Local model runtime and serving
• Cloud offload for larger models
• Easy switching between open families
• Private/offline experimentation

• gpt-oss
• Qwen3 / Qwen3.5 / Qwen3-Coder
• DeepSeek-R1 / DeepSeek-V4-Flash / DeepSeek-V4-Pro
• Gemma 3 / Gemma 3n
• Llama 4 / Llama 3.2 / Llama 3.1
• Phi-4 / Kimi K2.6 / gpt-oss-safeguard

• Local prototyping
• Privacy-first workflows
• Developer experimentation

The distribution and discovery layer for open models. It is the default place to watch releases, host weights, and track ecosystem momentum.

• Model hub and hosting
• Open-weight distribution
• Community evaluation and tooling

• Model discovery
• Open-model release tracking
• Community experimentation