DeepSeek R1
Product status: Status not recorded
No lifecycle review is recorded for this listing. Its inclusion in the directory does not confirm current availability; check the vendor before relying on it.
No status check date recorded
DeepSeek's original downloadable R1 reasoning model and related distilled checkpoints, with model-specific licenses and self-managed computing costs. No lifecycle review is recorded for this listing. Its inclusion in the directory does not confirm current availability; check the vendor before relying on it.
Product preview
Decision notes
Official-document research
Based on the cited documentation. We did not use the product, purchase a plan, or run quality, speed or reliability tests in this review.
Consider it for
- Teams managing model deployment and output review
Choose another approach if
- Assuming current DeepSeek Chat/API equals original R1
Pricing and conditions
Current terms unconfirmed
Original MIT weights/code do not require a license fee, but hardware, memory, storage, electricity and operations cost separately. The current API table prices deepseek-flash/V4 models, not a confirmed original R1 offer. No R1 currency/amount, region, tax, token tier/cache rate, quota or hosted contract was verified. The former V4 figures are withdrawn from this R1 entry; do not interpret the catalog's free classification as free API credits.
Pricing source read:
Pricing or terms sourceCode, model and service licensing
The card explicitly licenses original code and weights under MIT, permitting commercial use and derivatives with copyright/permission notices and no warranty. Distill-Qwen models derive from Apache-2.0 Qwen2.5; Distill-Llama 8B/70B derive from Llama3.1/3.3 with their respective base terms. Do not treat every R1-named fine-tune as identical MIT assets. Model permission does not clear input/output rights or establish a hosted service's training, logging or retention policy.
Basis and limits of the recorded score
The existing 4.6 score is retained; its original scoring evidence is unavailable. This documentation review did not measure performance or compare products and did not re-score it.
Overview
This entry covers the original DeepSeek-R1 release, not whatever model currently runs in DeepSeek Chat or a current API alias. Its card distinguishes R1, which uses cold-start data before reinforcement learning, from R1-Zero. R1 can generate reasoning text, but that text is not proof of correctness or a faithful explanation. The card gives decoding/prompt guidance and warns about repetition or incoherence. Select the exact full or distilled checkpoint and evaluate your task rather than carrying over vendor benchmark comparisons.
Sources
The factual claims in this review were read from the pages below. Each date is when an editor last opened the source, not when the product changed.
- Vendor documentation Checked 2026-10-15
- Vendor documentation Checked 2026-10-15
- Vendor pricing page Checked 2026-10-15
Pros / Cons
Pros
- Original model code/weights are published
- Checkpoint and prompting guidance
Cons
- Serving and evaluation require resources
- Distilled variants have different base terms
- Current API pricing does not establish an R1 quote
Who it's for
Reviewed reasoning or code experiments
Evaluating a specifically licensed checkpoint
Bottom line
Consider the original R1 weights when you can evaluate outputs and operate the required stack; confirm any hosted R1 offer separately.
Key Features
Tags
Compare DeepSeek R1 with alternatives
Frequently asked questions
Is DeepSeek R1 currently available?
No lifecycle review is recorded for this listing. Its inclusion in the directory does not confirm current availability; check the vendor before relying on it.
What is DeepSeek R1?
DeepSeek R1 is a Conversational AI tool developed by DeepSeek. DeepSeek's original downloadable R1 reasoning model and related distilled checkpoints, with model-specific licenses and self-managed computing costs.
What costs and terms should I check for DeepSeek R1?
Original MIT weights/code do not require a license fee, but hardware, memory, storage, electricity and operations cost separately. The current API table prices deepseek-flash/V4 models, not a confirmed original R1 offer. No R1 currency/amount, region, tax, token tier/cache rate, quota or hosted contract was verified. The former V4 figures are withdrawn from this R1 entry; do not interpret the catalog's free classification as free API credits.
Is DeepSeek R1 open source?
The card explicitly licenses original code and weights under MIT, permitting commercial use and derivatives with copyright/permission notices and no warranty. Distill-Qwen models derive from Apache-2.0 Qwen2.5; Distill-Llama 8B/70B derive from Llama3.1/3.3 with their respective base terms. Do not treat every R1-named fine-tune as identical MIT assets. Model permission does not clear input/output rights or establish a hosted service's training, logging or retention policy.
What are the best alternatives to DeepSeek R1?
Notable Conversational AI alternatives include ChatGPT, Claude, Gemini. Browse all 67 tools in our Conversational AI category for a full overview.
Was this listing useful?
Basic Info
- Pricing
- Free
Original MIT weights/code do not require a license fee, but hardware, memory, storage, electricity and operations cost separately. The current API table prices deepseek-flash/V4 models, not a confirmed original R1 offer. No R1 currency/amount, region, tax, token tier/cache rate, quota or hosted contract was verified. The former V4 figures are withdrawn from this R1 entry; do not interpret the catalog's free classification as free API credits.
- Editorial score
- ★ 4.6 / 5 How we score
- Platforms
- Self-hostedModel weights
- Last verified
- 2026-10-12
