{"count":19,"drivers":[{"id":"anthropic/claude-sonnet-4-6","vendor":"anthropic","family":"anthropic","name":"Claude Sonnet 4.6","version":"4.6","description":"Strongest Claude in the chain — Mind organ HARD-TURN escalation. 200k context, extended thinking, native tool use, vision read, prompt caching. 1M-ctx variant available via metadata flag.","capabilities":["stream","text.generate","tool.use","vision.read"],"context_tokens":200000,"max_output_tokens":8192,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.sse","endpoint_template":"https://api.anthropic.com/v1/messages","auth_method":"api_key_header","auth_secret_key":"ANTHROPIC_API_KEY","region":null,"pricing":{"in_per_million_usd":3.0,"out_per_million_usd":15.0,"per_image_usd":null,"per_second_usd":null,"notes":"3× Haiku per output token. Cache math identical (0.1× / 1.25× of input). Escalate from Haiku only when the cheap model refuses or fails to converge."},"rate_limits":{"rpm":500,"tpm":200000,"rpd":null,"concurrency":null},"health":{"method":"none","path":null,"timeout_ms":0},"adapter_dotted_path":"vikii.adapters.anthropic:AnthropicAdapter","tags":["extended-thinking","mind-escalation","prompt-caching","tool-use"],"deprecated":false,"probe_path":null},{"id":"google/gemini-3.5-flash","vendor":"google","family":"gemini","name":"Gemini 3.5 Flash","version":"3.5-flash","description":"Primary text + tools + vision workhorse on direct Gemini API. Frontier-class agentic + coding performance, native function calling, sub-second TTFT, ~5× cheaper than Pro on text.","capabilities":["stream","text.generate","tool.use","vision.read"],"context_tokens":1048576,"max_output_tokens":65536,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/gemini-3.5-flash:generateContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":0.3,"out_per_million_usd":2.5,"per_image_usd":null,"per_second_usd":null,"notes":"Google AI Studio list price; verify on ai.google.dev/pricing."},"rate_limits":{"rpm":1000,"tpm":null,"rpd":2000000,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/gemini-3.5-flash","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["fast","frontier","orchestrator","tool-use","workhorse"],"deprecated":false,"probe_path":"vikii/data/probes/gemini-3.5-flash.json"},{"id":"google/gemini-3.1-pro","vendor":"google","family":"gemini","name":"Gemini 3.1 Pro","version":"3.1-pro-preview","description":"Quality-tier text + tools + vision. Advanced intelligence + agentic + complex problem-solving. Orchestrate fallback + deep_think primary.","capabilities":["stream","text.generate","tool.use","vision.read"],"context_tokens":2097152,"max_output_tokens":65536,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/gemini-3.1-pro-preview:generateContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":1.25,"out_per_million_usd":10.0,"per_image_usd":null,"per_second_usd":null,"notes":"Preview pricing — verify on ai.google.dev/pricing."},"rate_limits":{"rpm":300,"tpm":null,"rpd":500000,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/gemini-3.1-pro-preview","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["deep-think","orchestrator","preview","quality","tool-use"],"deprecated":false,"probe_path":"vikii/data/probes/gemini-3.1-pro-preview.json"},{"id":"google/gemini-embedding-2","vendor":"google","family":"gemini","name":"Gemini Embedding 2","version":"2","description":"Multimodal embedding model — text, images, video, audio, and PDFs in one embedding space. Downcastable via output_dimensionality.","capabilities":["embed"],"context_tokens":null,"max_output_tokens":null,"embed_dim":3072,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/gemini-embedding-2:embedContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":0.025,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":""},"rate_limits":{"rpm":1500,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/gemini-embedding-2","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["hippocampus","multimodal","rag","vikii-default-embedder"],"deprecated":false,"probe_path":"vikii/data/probes/gemini-embedding-2.json"},{"id":"google/imagen-4-fast","vendor":"google","family":"imagen","name":"Imagen 4 Fast","version":"4.0","description":"Faster, cheaper Imagen 4 variant. ~3× cheaper than Ultra; ideal for volume image generation where iteration speed matters.","capabilities":["image.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/imagen-4.0-fast-generate-001:predict","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":0.02,"per_second_usd":null,"notes":""},"rate_limits":{"rpm":30,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/imagen-4.0-fast-generate-001","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["designer-volume","imagination-organ"],"deprecated":false,"probe_path":"vikii/data/probes/imagen-4-fast.json"},{"id":"google/imagen-4-ultra","vendor":"google","family":"imagen","name":"Imagen 4 Ultra","version":"4.0","description":"High-fidelity photoreal + design. Max prompt adherence in the Imagen 4 line; one image per request.","capabilities":["image.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/imagen-4.0-ultra-generate-001:predict","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":0.06,"per_second_usd":null,"notes":""},"rate_limits":{"rpm":20,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/imagen-4.0-ultra-generate-001","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["designer-first-draft","imagination-organ"],"deprecated":false,"probe_path":"vikii/data/probes/imagen-4-ultra.json"},{"id":"google/nano-banana-pro","vendor":"google","family":"gemini","name":"Nano Banana Pro","version":"nano-banana-pro-preview","description":"Conversational image edit + generate. Preserves identity across iterations; multimodal IMAGE+TEXT output via :generateContent.","capabilities":["image.edit","image.generate","text.generate","vision.read"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/nano-banana-pro-preview:generateContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":"Per-image-edit pricing — verify on Google's current pricing page (preview)."},"rate_limits":{"rpm":60,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/nano-banana-pro-preview","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["image-editor","imagination-organ"],"deprecated":false,"probe_path":"vikii/data/probes/nano-banana-pro.json"},{"id":"google/veo-3-fast","vendor":"google","family":"veo","name":"Veo 3 Fast","version":"3.1-fast","description":"Asynchronous video generation via :predictLongRunning. ~3× cheaper than full-quality Veo 3 at slight fidelity cost.","capabilities":["video.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/veo-3.1-fast-generate-preview:predictLongRunning","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":0.2,"notes":"Veo 3.1 Fast preview — ~3× cheaper than full Veo 3.1 (~$0.50-0.75/sec)."},"rate_limits":{"rpm":5,"tpm":null,"rpd":null,"concurrency":2},"health":{"method":"http_get","path":"/v1beta/models/veo-3.1-fast-generate-preview","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["imagination-organ","long-running","videographer"],"deprecated":false,"probe_path":"vikii/data/probes/veo-3-fast.json"},{"id":"google/veo-3-pro","vendor":"google","family":"veo","name":"Veo 3.1","version":"3.1","description":"THE single video model — Veo 3.1, the latest video model on the Gemini API. Asynchronous via :predictLongRunning. Native synced audio, up to 3 reference images, image-to-video, and clip extension.","capabilities":["video.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/veo-3.1-generate-preview:predictLongRunning","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":0.6,"notes":"Veo 3.1 — ~$0.50-0.75/sec."},"rate_limits":{"rpm":5,"tpm":null,"rpd":null,"concurrency":2},"health":{"method":"http_get","path":"/v1beta/models/veo-3.1-generate-preview","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["imagination-organ","long-running","videographer"],"deprecated":false,"probe_path":"vikii/data/probes/veo-3-fast.json"},{"id":"google/lyria-3-clip","vendor":"google","family":"lyria","name":"Lyria 3 Clip","version":"3-clip-preview","description":"Short Lyria 3 music clips (~30s) on direct Gemini API. Cheaper draft pass before Lyria 3 Pro for finals.","capabilities":["music.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":48000,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/lyria-3-clip-preview:generateContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":"Per-clip; cheaper than Lyria 3 Pro."},"rate_limits":{"rpm":5,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/lyria-3-clip-preview","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["composer-draft","imagination-organ","preview"],"deprecated":false,"probe_path":"vikii/data/probes/lyria-3-clip.json"},{"id":"google/lyria-3-pro","vendor":"google","family":"lyria","name":"Lyria 3 Pro","version":"3-pro-preview","description":"Full-length Lyria 3 (~3 min). Genre-faithful, instrument-controllable. Higher fidelity than Clip.","capabilities":["music.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":48000,"locality":"cloud","transport":"http.rest","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/lyria-3-pro-preview:generateContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":"Per-clip pricing; higher than Lyria 3 Clip."},"rate_limits":{"rpm":5,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":"/v1beta/models/lyria-3-pro-preview","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["composer","imagination-organ","preview"],"deprecated":false,"probe_path":"vikii/data/probes/lyria-3-pro.json"},{"id":"google/gemini-3.1-flash-live","vendor":"google","family":"gemini","name":"Gemini Live 2.5 Flash Native Audio","version":"2.5-flash-native-audio-latest","description":"Bidirectional voice agent via WebSocket. Native audio I/O — the model generates audio directly, no separate TTS stage. Preview channel.","capabilities":["audio.in","audio.out","live.bidi","stream","text.generate","tool.use"],"context_tokens":128000,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":24000,"locality":"cloud","transport":"websocket","endpoint_template":"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-native-audio-latest:bidiGenerateContent","auth_method":"api_key_header","auth_secret_key":"GEMINI_API_KEY","region":null,"pricing":{"in_per_million_usd":0.5,"out_per_million_usd":2.0,"per_image_usd":null,"per_second_usd":0.0006,"notes":"Live native-audio: text-token + per-second audio billing (preview)."},"rate_limits":{"rpm":60,"tpm":null,"rpd":null,"concurrency":8},"health":{"method":"http_get","path":"/v1beta/models/gemini-2.5-flash-native-audio-latest","timeout_ms":400},"adapter_dotted_path":"vikii.providers.google:GoogleProvider","tags":["bidi","live","native-audio","preview","voice"],"deprecated":false,"probe_path":"vikii/data/probes/gemini-3.1-flash-live.json"},{"id":"local/bge-reranker-base","vendor":"local","family":"bge","name":"BGE Reranker Base","version":"1.0","description":"BAAI/bge-reranker-base via onnxruntime. Cross-encoder for RAG candidate rerank. Server-side local.","capabilities":["embed"],"context_tokens":null,"max_output_tokens":null,"embed_dim":0,"audio_sample_rate_hz":null,"locality":"region","transport":"local.process","endpoint_template":"","auth_method":"none","auth_secret_key":"","region":null,"pricing":{"in_per_million_usd":0.0,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":""},"rate_limits":{"rpm":null,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":null,"timeout_ms":200},"adapter_dotted_path":"vikii.adapters.local:LocalRerankAdapter","tags":["offline-first","rag","rerank"],"deprecated":false,"probe_path":null},{"id":"local/multilingual-e5-base","vendor":"local","family":"e5","name":"Multilingual E5 Base","version":"1.0","description":"intfloat/multilingual-e5-base via fastembed (ONNX). 768-dim native, 100 languages, MTEB-leading in base size. Server-side local.","capabilities":["embed"],"context_tokens":null,"max_output_tokens":null,"embed_dim":768,"audio_sample_rate_hz":null,"locality":"region","transport":"local.process","endpoint_template":"","auth_method":"none","auth_secret_key":"","region":null,"pricing":{"in_per_million_usd":0.0,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":""},"rate_limits":{"rpm":null,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":null,"timeout_ms":200},"adapter_dotted_path":"vikii.adapters.local:LocalEmbedAdapter","tags":["hippocampus","local-embedder","offline-first","rag"],"deprecated":false,"probe_path":null},{"id":"local/nllb-200-distilled","vendor":"local","family":"nllb","name":"NLLB-200 Distilled","version":"600M","description":"Meta NLLB-200 distilled-600M via transformers. 200 languages, translation source-of-truth for the offline-first stack.","capabilities":["text.generate"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"region","transport":"local.process","endpoint_template":"","auth_method":"none","auth_secret_key":"","region":null,"pricing":{"in_per_million_usd":0.0,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":null,"notes":""},"rate_limits":{"rpm":null,"tpm":null,"rpd":null,"concurrency":null},"health":{"method":"http_get","path":null,"timeout_ms":200},"adapter_dotted_path":"vikii.adapters.local:LocalTranslationAdapter","tags":["nllb","offline-first","translation"],"deprecated":false,"probe_path":null},{"id":"replicate/wav2lip","vendor":"replicate","family":"wav2lip","name":"Wav2Lip on Replicate","version":"cjwbw/wav2lip:8d65e3f4f4298520e079198b493c25adfc43c058ffec924f2aefc8010ed25eef","description":"Third-party Wav2Lip lip-sync via Replicate's prediction API. Takes a face image / video clip + a voice WAV and returns an lip-synced mp4. Stage 3 / decision D2=B — we never host the Wav2Lip checkpoint, every call hits Replicate.","capabilities":["lip.sync"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://api.replicate.com/v1/predictions","auth_method":"bearer","auth_secret_key":"REPLICATE_API_TOKEN","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":0.0023,"notes":"Replicate billing is per CPU/GPU-second on the model. Indicative rate; ops should reconcile against actual invoice."},"rate_limits":{"rpm":10,"tpm":null,"rpd":null,"concurrency":4},"health":{"method":"http_get","path":"/v1/models/cjwbw/wav2lip","timeout_ms":600},"adapter_dotted_path":"vikii.services.lab.production.replicate_lipsync:ReplicateLipsyncProvider","tags":["lab-only","lip-sync","replicate","third-party"],"deprecated":false,"probe_path":null},{"id":"replicate/codeformer","vendor":"replicate","family":"codeformer","name":"CodeFormer (face restore)","version":"sczhou/codeformer:7de2ea26c616d5bf2245ad0d5e24f0ff9a6204578a5c876db53142edd9d2cd56","description":"Third-party face restoration via Replicate. Restores blurry / low-resolution / damaged faces while preserving identity. Used by the Photo Studio Restore mode.","capabilities":["image.face_restore","image.restore"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://api.replicate.com/v1/predictions","auth_method":"bearer","auth_secret_key":"REPLICATE_API_TOKEN","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":0.0023,"notes":"Replicate billing is per CPU/GPU-second on the model. Indicative rate; reconcile against the actual invoice."},"rate_limits":{"rpm":20,"tpm":null,"rpd":null,"concurrency":2},"health":{"method":"http_get","path":"/v1/models/sczhou/codeformer","timeout_ms":600},"adapter_dotted_path":"vikii.services.lab.replicate_http:ReplicateImageProvider","tags":["lab-only","photo-studio","replicate","third-party"],"deprecated":false,"probe_path":null},{"id":"replicate/real-esrgan","vendor":"replicate","family":"real-esrgan","name":"Real-ESRGAN (super-resolution)","version":"nightmareai/real-esrgan:f121d640bd286e1fdc67f9799164c1d5be36ff74576ee11c803ae5b665dd46aa","description":"Third-party super-resolution via Replicate. Upscales 2× / 4× with practical denoising; used by the Photo Studio Restore stage and by every mode's HD / 2K / 4K export.","capabilities":["image.upscale"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://api.replicate.com/v1/predictions","auth_method":"bearer","auth_secret_key":"REPLICATE_API_TOKEN","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":0.0023,"notes":"Replicate billing is per CPU/GPU-second on the model. Indicative rate; reconcile against the actual invoice."},"rate_limits":{"rpm":20,"tpm":null,"rpd":null,"concurrency":2},"health":{"method":"http_get","path":"/v1/models/nightmareai/real-esrgan","timeout_ms":600},"adapter_dotted_path":"vikii.services.lab.replicate_http:ReplicateImageProvider","tags":["lab-only","photo-studio","replicate","third-party"],"deprecated":false,"probe_path":null},{"id":"replicate/bringing-old-photos-back","vendor":"replicate","family":"bopbl","name":"Bringing Old Photos Back to Life","version":"microsoft/bringing-old-photos-back-to-life:c75db81db6cbd809d93cc3b7e7a088a351a3349c9fa02b6c0810da7c8b1bbb1c","description":"Third-party scratch / tear removal and missing-region inpaint for old photographs. Microsoft's BOPBL reference model hosted on Replicate. Used by the Photo Studio Restore mode when the damage_analysis pass detects scratches or torn regions.","capabilities":["image.restore"],"context_tokens":null,"max_output_tokens":null,"embed_dim":null,"audio_sample_rate_hz":null,"locality":"cloud","transport":"http.rest","endpoint_template":"https://api.replicate.com/v1/predictions","auth_method":"bearer","auth_secret_key":"REPLICATE_API_TOKEN","region":null,"pricing":{"in_per_million_usd":null,"out_per_million_usd":null,"per_image_usd":null,"per_second_usd":0.0023,"notes":"Replicate billing is per CPU/GPU-second on the model. Indicative rate; reconcile against the actual invoice."},"rate_limits":{"rpm":15,"tpm":null,"rpd":null,"concurrency":2},"health":{"method":"http_get","path":"/v1/models/microsoft/bringing-old-photos-back-to-life","timeout_ms":600},"adapter_dotted_path":"vikii.services.lab.replicate_http:ReplicateImageProvider","tags":["lab-only","photo-studio","replicate","third-party"],"deprecated":false,"probe_path":null}],"filters":{"capability":null,"vendor":null}}