Audio · karta decyzji
Inkling
Thinking Machines Lab
Inkling is a 975B parameter open-weights multimodal AI model from Thinking Machines Lab, supporting text, image, and audio input with a 1M token context window. It enables fine-tuning via Tinker and inference via API partners.
Dostęp
Do weryfikacji
Cena od
4.05
Trial
unknown
Sprawdzone
9.08.2026
W skrócie
Czy to narzędzie pasuje do Twojego zadania?
The model weights are free under Apache 2.0. Inference via API partners (e.g., Together AI, Baseten) is billed per token at $4.05 per 1,000 tokens.
Zastosowania
- research
- content generation
- SEO
- automation
- transcription
- code generation
- infrastructure
Najważniejsze funkcjonalności
- multimodal input (text, images, audio)
- output in text and code
- context window of up to 1 million tokens
- adjustable reasoning effort (fine-tuning of thinking depth)
- Mixture of Experts (MoE) architecture
- support for tool calling
- structured output
- available in multiple quantization formats (GGUF, BF16, NVFP4)
- fine-tuning available via Tinker
- open weights under Apache 2.0 license
Darmowy plan i trial
Darmowy dostęp: Yes, the model weights are freely downloadable on Hugging Face under the Apache 2.0 license.
Trial: unknown
Cennik: The model weights are free under Apache 2.0. Inference via API partners (e.g., Together AI, Baseten) is billed per token at $4.05 per 1,000 tokens.
Plany i limity
| Plan | Cena | Dostęp | Limit / zawartość | Trial / karta |
|---|---|---|---|---|
| Inference via API partners | 4.05 USD / per_token | open_weights | unknown | unknown · karta: no |
Dane i bezpieczeństwo
- Prywatność
- unknown
- Trening na danych
- unknown
- Region danych
- unknown
- RODO / GDPR
- unknown
- Bezpieczeństwo
- unknown
Materiały i dowody
unknown
Materiały producenta
Tytuł materiału producenta nie jest niezależnym dowodem skuteczności. Źródła oznaczamy według ich pochodzenia.
Mocne strony
- + Open weights under permissive Apache 2.0 license
- + Multimodal input (text, images, audio)
- + Extremely large context window (1M tokens)
- + Adjustable reasoning effort for cost control
- + Fine-tuning available via Tinker
- + Supports tool calling and structured output
Na co uważać
- Outputs are limited to text and code; no image generation
- Performance is lower than top-tier proprietary models
- Full model requires large GPU clusters
- Only available in specific quantized formats for local inference