Models stay visible while routes come online. Available routes have an active upstream key mapped to that exact model; everything else is marked as coming soon.
89 of 89 models · 27 available · 66 coming soon
sandbase-claude-opus-4-8
Claude Opus 4.8 through SandBase OpenAI-compatible routing. 1M context, 128k output.
bedrock-claude-opus-4-8
gemini-2.5-flash-tts
Cheaper prior-gen TTS — same controllability, narrower language set.
gemini-3-flash
Mid-tier multimodal. Image + video + text input. 1M context.
gemini-3.1-flash-live
Realtime conversational audio — bidirectional, sub-second latency. ~$0.005/min in, $0.018/min out.
gemini-3.1-flash-tts
Controllable TTS across 70+ languages, 200+ inline emotion tags. Audio output tokens.
gemini-3.1-flash-lite
Cheapest 1M-context tier. Great for high-volume agents.
gemini-3.1-pro
Google’s flagship reasoning model + Computer Use. 1M context.
vertex-gemini-3.1-pro
gemini-3.5-flash
Stable Gemini 3.5 model built for fast frontier intelligence, agents, coding, and grounding.
nano-banana
Original Nano Banana (Gemini 2.5 Flash Image). Fast, fun edits.
nano-banana-2
Nano Banana 2 (Gemini 3.1 Flash Image) — Pro features at Flash speed. $0.067/1K, $0.10/2K, $0.15/4K.
nano-banana-pro
Nano Banana Pro (Gemini 3 Pro Image) — best fidelity, complex prompts, accurate text. $0.134/1K-2K, $0.24/4K.
veo-3.1
Cinematic text/image-to-video, 720p–1080p, optional audio.
vertex-veo-3.1
veo-3.1-fast
Veo 3.1 Fast — $0.10/s @720p, $0.12/s @1080p, $0.30/s @4K.
vertex-veo-3.1-fast
bedrock-claude-haiku-4-5
Claude Haiku 4.5 through AWS Bedrock Global application inference profile.
bedrock-claude-opus-4-1-us-cr
Claude Opus 4.1 through AWS Bedrock US cross-region profile. Includes the documented 10% premium.
bedrock-claude-opus-4-5
Claude Opus 4.5 through AWS Bedrock Global application inference profile.
bedrock-claude-opus-4-6
Claude Opus 4.6 through AWS Bedrock Global application inference profile.
bedrock-claude-opus-4-7
Claude Opus 4.7 through AWS Bedrock Global application inference profile.
bedrock-claude-sonnet-4-5
Claude Sonnet 4.5 through AWS Bedrock Global application inference profile.
bedrock-claude-sonnet-4-6
Claude Sonnet 4.6 through AWS Bedrock Global application inference profile.
vertex-chirp-3
TTS with Instant Custom Voice (10s reference audio).
vertex-imagen-3
Highest-quality Imagen text-to-image with inpainting + editing.
vertex-lyria-3-pro
Music generation up to 184s. Public preview.
claude-fable-5
Claude Fable 5. Highest quality long-context Claude tier with a data-use notice required before playground use.
gpt-image-2
OpenAI image gen — flexible sizes, high-fidelity image input.
gpt-realtime
Bidirectional voice + text. ~$0.06/min audio in, $0.24/min audio out. Text in/out at $5/$20.
gpt-4o-audio-preview
Text-or-audio in, text-or-audio out via /v1/chat/completions.
gpt-4o-mini-tts
Token-priced TTS — ~$0.015/min generated audio.
gpt-4o-transcribe
ASR with optional speaker diarization. Same price as Whisper.
gpt-5.4
March 2026 release — sits between mini and 5.5. Strong cost/intel ratio.
gpt-5.4-mini
GPT-5.4-class capability, fast + efficient. 400k context.
gpt-5.4-nano
OpenAI’s smallest, cheapest model. High-volume simple tasks.
gpt-5.5
Flagship multimodal. 1M context. Computer use, MCP, hosted shell.
gpt-5.5-pro
Highest-tier flagship — research-grade reasoning. No cached-input discount.
o4-mini
Cheap reasoning model — math, code, structured analysis at low cost.
text-embedding-3-large
OpenAI large embedding model with higher retrieval quality.
text-embedding-3-small
OpenAI small embedding model for semantic search and clustering.
tts-1
Standard TTS — same voices, lower fidelity, half the cost of HD.
tts-1-hd
High-fidelity TTS — 6 preset voices. /v1/audio/speech endpoint.
whisper-1
Speech-to-text (ASR). Will be superseded by gpt-4o-transcribe.
minimax-hailuo-2.3
Latest Hailuo text-to-video. Billed in units; ~$0.04/s @768p.
minimax-hailuo-2.3-fast
Hailuo 2.3 Fast — 30% cheaper than 2.3, same family.
minimax-image-01
Text-to-image, photoreal + stylized.
minimax-m2.1
Older M-series — kept for reproducibility.
minimax-m2.5
Prior-gen M-series — same base price, slightly weaker quality.
minimax-m2.7
MiniMax’s newest self-iterating flagship for code + agents.
minimax-m2.7-highspeed
M2.7 with priority routing — 2× cost for lower TTFT.
minimax-music-1.5
Music generation from prompt — vocals + instrumentation.
minimax-speech-02-hd
Prior-gen HD voice — kept for reproducibility of pipelines.
minimax-speech-2.5-turbo
HD TTS — 40 languages, accurate voice replication.
minimax-speech-2.6
Latest TTS — Fluent LoRA voice cloning, prosodic naturalness across 40+ languages.
kimi-k2-0711-preview
July 2025 K2 preview snapshot — scheduled for discontinuation 2026-05-25. (Not currently listed on OpenRouter — keeping prior estimate.)
kimi-k2-0905-preview
September 2025 K2 preview snapshot — scheduled for discontinuation 2026-05-25.
kimi-k2-thinking
Explicit thinking-mode K2 — chain-of-thought reasoning for math + code.
kimi-k2-turbo
Moonshot’s agentic K2 with priority routing — fast TTFT, 128k context.
kimi-k2.5
January 2026 multimodal release — cheaper than K2.6 with similar capability.
kimi-k2.6
Moonshot’s April 2026 flagship — multimodal, 256k context.
kimi-latest
Auto-routes to Moonshot’s current default (K2.6 as of 2026-05). Multimodal.
moonshot-v1-128k
Legacy text-only Moonshot v1 long-context. (Not on OpenRouter — keeping prior flat pricing.)
deepseek-chat
Legacy alias → V4 Flash non-thinking mode. Sunsets 2026-07-24; migrate to deepseek-v4-flash.
deepseek-reasoner
Legacy alias → V4 Flash thinking mode. Sunsets 2026-07-24; migrate to deepseek-v4-flash with reasoning. (Not separately listed on OpenRouter — mirrors v4-flash.)
deepseek-v4-flash
V4 default — 1M context, 384k max output.
deepseek-v4-pro
V4 flagship reasoning — 1M context. OR pricing reflects 75%-off promo through 2026-05-31.
cogvideox
Zhipu’s text/image-to-video — 6s clips at 720p–1080p.
cogvideox-flash
Free video generation tier.
cogview-3-flash
Free image generation tier — generous rate limits.
cogview-4
Latest text-to-image — strong on Chinese text rendering + complex prompts.
glm-4-airx
Low-latency AirX variant — fastest GLM tier, smaller context.
glm-4-flash
Free tier — high-volume, simple tasks. Generous rate limits.
glm-4.6
GLM 4 family flagship — 200k context, agentic + coding focused.
glm-4.6v
GLM 4.6 vision variant — multimodal input on the 4-family flagship.
glm-4.7
Previous-gen flagship before GLM 5 — solid coding + reasoning at lower cost.
glm-4.7-flash
Lowest tier — high-volume simple tasks. Generous rate limits.
glm-5
GLM 5 base flagship — released 2026-02-12, available on Pro and Max coding tiers.
glm-5-turbo
March 2026 turbo variant — priority routing on the GLM 5 family.
glm-5.1
Zhipu’s April 2026 flagship — SOTA on SWE-Bench Pro, 200k context, 128k max output.
glm-5v-turbo
GLM 5 family vision-capable turbo — multimodal with priority routing.
glm-realtime
End-to-end voice + video understanding with singing + 2-min memory. Function calls supported.
glm-tts
Controllable + emotion-expressive zero-shot voice cloning. Open-sourced Dec 2025.
grok-4.1-fast
Budget-friendly Grok — low TTFT, very cheap for high-volume tasks.
grok-4.20
Long-context variant — 2M context window for very large corpora.
grok-4.3
xAI’s May 2026 flagship — 1M context, multimodal.
qwen3-asr-flash
Speech-to-text — multilingual, low-latency.
qwen3-max
Flagship Qwen3 — Alibaba’s largest dense + MoE model.
qwen3-omni-flash
Omni-modal: text + image + audio in/out. Cheapest realtime-class tier.
qwen3-vl-plus
Vision-language flagship — strong OCR, document QA, GUI grounding.
qwen3.5
Feb 2026 release — 397B params, 201 languages, 19× faster than prior gen.
qwen3.5-0.8b
Tiniest Qwen — extreme high-volume / on-device fallback.
cohere-rerank-v3.5
Cohere reranker for search and retrieval ordering via /v1/rerank.