Fournisseurs d’IA : comparatif

DeepInfra vs Together AI

Prix, fonctionnalités documentées et limites issus de sources officielles datées, côte à côte. Les inconnues restent inconnues ; des unités de facturation différentes ne permettent pas de comparer directement les coûts.

Ce que les sources établissent

Une lecture de la documentation officielle vérifiée pour DeepInfra et Together AI. Ce n’est ni un test pratique, ni un classement, ni la recommandation d’un produit.

Questions d’achat traitées pour les deux
5 critères
Documenté pour un seul
0 critère
Non vérifié pour aucun
0 critère
Prix
Pas encore comparables

Prix

Aucune comparaison de prix complète à conditions égales n’est encore établie ; les prochaines étapes indiquent ce qui manque.

Questions d’achat traitées pour les deux

Même quand les libellés concordent, vérifiez dans chaque source les différences d’offres, d’unités et de périmètre.

Interfaces API
DeepInfra An OpenAI-compatible API at https://api.deepinfra.com/v1/openai; the documentation says an existing OpenAI SDK can be pointed at it without code changes.
Together AI OpenAI-compatible serverless text generation
Catalogue de modèles
DeepInfra 100+ LLMs through an OpenAI-compatible API; Vision and OCR models; Embedding and reranker models; Image and video generation (FLUX, Stable Diffusion); Speech recognition (Whisper) and text-to-speech models
Together AI Kimi K3, GLM-5.3, GLM-5.3-Flash
Base de facturation
DeepInfra Some language models are billed per token and most other models by inference execution time; the pricing page says there are no long-term contracts or upfront costs.
Together AI Models are priced per 1 million tokens, with separate input and output prices for each model.
Contrôles des données
DeepInfra Input data is not stored to disk during inference and outputs are not stored; DeepInfra says it does not train on or share submitted data, with exceptions for Google and Anthropic models.
Together AI By default prompts and responses are stored and may be used for product improvements; organization admins can turn storage off to enable zero data retention.
Disponibilité
DeepInfra An account needs a card or pre-payment before the services can be used, and each account is limited to 200 concurrent requests.
Together AI Serverless performance is best-effort and requests may be limited under high demand; committed throughput with reliability guarantees is offered separately as provisioned throughput.

Limites à vérifier

  • DeepInfra The captured client README describes authentication using a DeepInfra API key.
  • Together AI The publisher describes Python SDK v2 examples for Together AI.

Pas encore vérifié

Chaque critère comparé est documenté pour au moins un produit.

« Non vérifié » signifie que nos sources vérifiées ne l’établissent pas. Cela ne veut pas dire qu’un produit n’a pas la fonctionnalité ni qu’il est gratuit.

Preuves côte à côte

Valeurs documentées avec leurs sources et dates de vérification ; les exigences que les sources n’établissent pas restent signalées.

  • Documenté comme pris en charge
  • Documenté comme non pris en charge
  • Non vérifié Les sources ne l’établissent pas
  • Sans objet
Périmètre et limites documentés ; les critères non vérifiés restent inconnus.
ExigenceDeepInfraTogether AI
Tarifs
Tarification API à l’usage
Tarification API à l’usage
Pour qui
The DeepInfra Python repository describes a client for inference and sandbox operations.

Vérifié le 4 nov. 2026 · DeepInfra: deepinfra-api-integration-source

The publisher documents serverless text generation through Together AI.

Vérifié le 3 nov. 2026 · Together AI: together-docs

Limites de couverture
The captured client README describes authentication using a DeepInfra API key.

Vérifié le 4 nov. 2026 · DeepInfra: deepinfra-api-integration-source

The publisher describes Python SDK v2 examples for Together AI.

Vérifié le 3 nov. 2026 · Together AI: together-docs

Interfaces API
An OpenAI-compatible API at https://api.deepinfra.com/v1/openai; the documentation says an existing OpenAI SDK can be pointed at it without code changes.

DeepInfra documentation start page captured 9 November 2026. Compatibility is the supplier's statement; individual models and features may differ and no request was made.

Vérifié le 9 nov. 2026 · DeepInfra: deepinfra-api-docs-20261109b

OpenAI-compatible serverless text generation

Publisher skills README (togethercomputer/skills) captured 3 November 2026; supplier description of the chat-completions skill, not an API call or compatibility test. Which OpenAI endpoints and parameters are supported is not established.

Vérifié le 11 nov. 2026 · Together AI: together-docs

Catalogue de modèles
100+ LLMs through an OpenAI-compatible API; Vision and OCR models; Embedding and reranker models; Image and video generation (FLUX, Stable Diffusion); Speech recognition (Whisper) and text-to-speech models

DeepInfra documentation start page captured 9 November 2026. Categories of models as the page describes them, not a complete or current model list; availability and rates differ per model. Supplier statement, not a hands-on test.

Vérifié le 9 nov. 2026 · DeepInfra: deepinfra-api-docs-20261109b

Kimi K3, GLM-5.3, GLM-5.3-Flash

Together AI pricing page captured 9 November 2026 (amounts shown with a $ sign; the page states no ISO currency); supplier statement, not a hands-on test. Three of the serverless chat models listed in the pricing table; the table lists more and changes often, so it is not the complete catalogue.

Vérifié le 11 nov. 2026 · Together AI: together-ai-pricing-20261109b

Base de facturation
Some language models are billed per token and most other models by inference execution time; the pricing page says there are no long-term contracts or upfront costs.

DeepInfra pricing page captured 9 November 2026. Rates differ by model and are listed per model on the page; dedicated GPU deployments and embeddings have separate rate tables. The same page separately requires a card or pre-payment before use. Supplier statement, not a hands-on test; amounts are not restated here because they change.

Vérifié le 9 nov. 2026 · DeepInfra: deepinfra-api-pricing-20261109b

Models are priced per 1 million tokens, with separate input and output prices for each model.

Together AI pricing page captured 9 November 2026 (amounts shown with a $ sign; the page states no ISO currency); supplier statement, not a hands-on test. Serverless Inference pricing table (Chat tab): rates vary by model, some models list a cached-input price, and the table header also lists a Batch API price. Image, video, audio and other modalities use other units. Dedicated endpoints, provisioned throughput and GPU clusters are priced separately and not captured.

Vérifié le 11 nov. 2026 · Together AI: together-ai-pricing-20261109b

Contrôles des données
Input data is not stored to disk during inference and outputs are not stored; DeepInfra says it does not train on or share submitted data, with exceptions for Google and Anthropic models.

DeepInfra Data Privacy documentation captured 9 November 2026. The page also says image-generation outputs are stored briefly and that bulk inference APIs may store data longer in encrypted form. Supplier statement, not an audit or certification.

Vérifié le 9 nov. 2026 · DeepInfra: deepinfra-api-data-20261109b

By default prompts and responses are stored and may be used for product improvements; organization admins can turn storage off to enable zero data retention.

Together AI Privacy and security documentation captured 9 November 2026; supplier statement, not an audit or test. The page also says training on customer data is opt-in and off by default, and that some passthrough models forward prompts to third-party providers who apply their own data policies.

Vérifié le 11 nov. 2026 · Together AI: together-ai-privacy-and-security-20261109b

Disponibilité
An account needs a card or pre-payment before the services can be used, and each account is limited to 200 concurrent requests.

DeepInfra pricing page captured 9 November 2026. The page says pre-paid accounts draw down their balance, invoiced accounts are billed once at the beginning of the following month, a spending limit can be set, and the page invites contacting DeepInfra for more than 200 concurrent requests. Supplier statement, not a hands-on test.

Vérifié le 9 nov. 2026 · DeepInfra: deepinfra-api-pricing-20261109b

Serverless performance is best-effort and requests may be limited under high demand; committed throughput with reliability guarantees is offered separately as provisioned throughput.

Together AI serverless rate-limits documentation captured 9 November 2026; supplier statement, not a load test. Numeric limits are not stated on the page; the page says custom limits or access restrictions can apply to models in high demand.

Vérifié le 11 nov. 2026 · Together AI: together-ai-rate-limits-20261109b

Note de couverture

Integration README at the captured repository revision, 4 November 2026. Client-code licence and hosted service, model and trademark terms are separate. No account, model call, deployment or benchmark.

Publisher documentation captured 3 November 2026; describes selected documentation or client behavior, not a live service test. Current model availability, prices and service terms require confirmation.

Prochaines étapes

  • Choisissez d’abord le modèle exact de DeepInfra : ses tarifs par token dépendent du modèle et du niveau de service.
  • Choisissez d’abord le modèle exact de Together AI : ses tarifs par token dépendent du modèle et du niveau de service.

Les commissions ne modifient jamais les coûts, l’éligibilité ni l’ordre.