
Cohere Startup Credits: $2,000 in API credits
Enterprise AI models for text generation, embedding, classification, and reranking, alternative to OpenAI for NLP.
Premium deal
Already have an account? Log in
Deal Highlights
Our Cohere listing: $2,000 in API credits
Offer: $2,000 in API credits. Our listing's requirement is "AI startup building NLP features". There is no promo code. The link in our listing opens Cohere's June 2024 blog post announcing its startup program. It is not a SaaSOffers tracking link.
We checked these plans and prices against Cohere's official pricing and help pages in November 2026.
Cohere sells language models through an API: Command for text generation, Embed for turning text and images into vectors, and Rerank for sorting search results by relevance. Most of its marketing is aimed at enterprises, but the API is open to anyone who signs up.

The $2,000 figure is our listing's term. The blog post it links to is dated June 6, 2024, about 29 months before this check. It says the program helps early-stage companies use AI "at an affordable cost", but it gives no credit amount, no eligibility rules and no deadline. Cohere's pricing page and its FAQ in the docs do not mention a startup program at all. Treat $2,000 as unconfirmed until Cohere confirms it in writing, and ask what the credits cover and when they expire.
The free trial key is real, but it cannot run a product
Every Cohere account gets a Trial API key automatically. Cohere's pricing page says calls made with a Trial key are free, rate limited, and "not permitted to be used for production or commercial purposes". Cohere's docs say anyone can sign up and use all models and APIs with a trial key.
The docs add a hard monthly cap: trial keys are limited to 1,000 API calls a month. Production keys carry the same 1,000-call cap on newer Chat model variants such as Command A Reasoning, and the docs tell you to contact sales@cohere.com before using those in production.
A single page load that embeds a query, reranks 20 results and generates an answer uses three calls, so 1,000 calls is roughly 333 of those page loads for the whole month.
Production access is a separate step. Cohere says you need Owner privileges, then you open Billing and Usage in the dashboard, click "Get Your Production key" and fill out a Go to Production form. New accounts start as personal accounts. Organizations are not automatically approved for production keys.
Billing is pay as you go. Cohere issues a bill at the end of each calendar month, or sooner when your unpaid balance reaches $250.
Embed 5: $0.12 or $0.08 per million text tokens
Cohere launched Embed 5 on September 30, 2026. It comes in two tiers that share one embedding space, so you can index documents with Pro and query with Fast without rebuilding the index. Cohere says cross-tier search loses on average 1.6% for Fast queries and 2.7% for Pro queries against same-model baselines, and recommends indexing with Pro and querying with Fast.
| Tier | Text, per 1M tokens | Image, per 1M tokens | Speed (Cohere's figure) |
|---|---|---|---|
| Embed 5 Pro | $0.12 | $0.40 | 159.7 documents per second |
| Embed 5 Fast | $0.08 | $0.40 | 377.3 documents per second |
Both tiers take text, images and fused text plus image input. They have a 128K token context, more than 100 languages, output sizes of 2048, 1536, 1024, 768, 512 or 256 dimensions, and float, int8 or binary formats. Cohere says both can be self-hosted with vLLM.
Cohere's own benchmark table has Embed 5 Pro at 85.8 on ViDoRe V3, ahead of Voyage 4 Large at 83.7, Gemini Embedding 2 at 83.2 and OpenAI text-embedding-3-large at 75.5. These are Cohere's numbers, run by Cohere. Test on your own documents before you trust them.
Rerank 5: billed per search
Cohere released Rerank 5 on October 21, 2026. Cohere's pricing page defines a search unit as one query with up to 100 documents to be ranked. Any document over 500 tokens, query included, is split into chunks, and each chunk counts as a separate document. Cohere's docs say rerank models are "priced based on the quantity of searches". Cohere's pricing page shows no Rerank dollar price, so ask Cohere before you forecast.
| Model | Context | Notes |
|---|---|---|
| Rerank 5 Pro | 32,768 tokens (query plus document) | About 2B parameters, highest quality |
| Rerank 5 Fast | 32,768 tokens | About 500M parameters, 2.9x Pro's throughput per Cohere |
Rerank 5 returns a calibrated relevance score from 0 to 1 for each document. A feature called Context Focus is off by default. When you turn it on with relevant_spans, the API returns scored text spans inside each document, so you can send only the useful passages to your generator. Cohere says removing half the retrieved context kept 96% of answer quality, and removing three quarters kept 87%.
Rerank 5 takes text only and does not support instructions that change how relevance is judged.
Command: the pricing page lists only legacy models
Cohere's pricing page in November 2026 lists per-token prices only for legacy models, labeled for existing customers:
| Model | Input per 1M tokens | Output per 1M tokens |
|---|---|---|
| Command | $1.00 | $2.00 |
| Command-light | $0.30 | $0.60 |
| Command R (03-2024) | $0.50 | $1.50 |
| Command R+ (04-2024) | $3.00 | $15.00 |
| Command R+ (08-2024) | $2.50 | $10.00 |
| Aya Expanse 8B and 32B | $0.50 | $1.50 |
The current Command models are Command A+, Command A Reasoning, Command A Translate, Command A Vision, Command A, Command R+, Command R and Command R7B. The pricing page shows no price for any of them, and Cohere's docs say the pricing page holds the up-to-date prices. We could not verify a current per-token price for Command A or Command A+. Get a quote from Cohere before you commit to a generation budget.
Command A+ launched on May 20, 2026. Cohere's docs describe it as a sparse mixture-of-experts model with 218B total and 25B active parameters, 48 languages, vision and text input, and an optional thinking mode. It is released under Apache 2.0 on Hugging Face and runs on one B200 at W4A4 or two H100s at W4A4. Command A has 111B parameters, a 256K context window and 23 languages.

Rate limits by key type
Cohere's rate limit page gives these per-minute caps. Trial keys and production keys differ sharply on search and chat.
| Endpoint | Trial key | Production key |
|---|---|---|
| Chat, Command A, R+, R, R7B | 20 requests/min | 500 requests/min |
| Chat, Command A+, A Reasoning, A Translate, A Vision | 20 requests/min | Contact sales |
| Embed (text) | 2,000 inputs/min | 2,000 inputs/min |
| Embed (images) | 5 inputs/min | 400 inputs/min |
| EmbedJob | 5 requests/min | 50 requests/min |
| Rerank | 10 requests/min | 1,000 requests/min |
| Audio transcription | 5 requests/min | Contact sales |
| Tokenize | 100 requests/min | 2,000 requests/min |
To raise limits, the docs point you to support@cohere.com.
What $2,000 buys, if our listing's credit applies
These examples use only prices from Cohere's own pages. If our $2,000 term applies, the credits stretch like this.
Embedding a document library
Say you have 1 million support articles of 500 tokens each, which is 500 million tokens.
- Embed 5 Pro at $0.12 per 1M: 500 x $0.12 = $60
- Embed 5 Fast at $0.08 per 1M: 500 x $0.08 = $40
At $0.12 per million, $2,000 covers about 16.7 billion text tokens. At $0.08 it covers 25 billion.
Generating answers on a legacy model
One request with 2,000 tokens in and 500 out on Command R+ (08-2024) costs 2,000 x $2.50 / 1M = $0.005 plus 500 x $10.00 / 1M = $0.005, so $0.01. That is 200,000 requests for $2,000. On Command R (03-2024) it costs $0.001 plus $0.00075, so $0.00175, or about 1.14 million requests. These are legacy models for existing customers, so a new account may be steered to a newer one with a different price.
Deploying on your own hardware or in your own cloud
Cohere's FAQ says all of its models can be deployed privately, and you start by contacting sales. It also lists AWS Bedrock, AWS SageMaker, Azure AI and OCI Generative AI as places to reach the models. Embed 5 and Rerank 5 are also on Microsoft Foundry and SageMaker.
One licensing detail to read carefully: the FAQ says the default license for open weights is non-commercial, and commercial use needs a license from sales. Command A+ is the exception we found, because its docs state Apache 2.0. Check the license on the exact model card you download.
Model Vault, the dedicated option
Model Vault runs a model on dedicated GPUs, billed per GPU hour. Cohere's page says pricing depends on the GPU and the number of replicas.
| GPU | On demand, per GPU hour | Promotional, through March 31, 2027 |
|---|---|---|
| L4 24GB | $3.00 | $1.00 |
| H100 80GB | $6.25 | $4.25 |
| H100 80GB encrypted | $9.00 | $4.25 |
| B200 192GB | $9.75 | $8.00 |
| B200 192GB encrypted | $12.00 | $8.00 |
The encrypted rows use Confidential Compute. One H100 running around the clock for 30 days is 720 hours, which is $4,500 on demand or $3,060 at the promotional rate. That is more than the whole $2,000 credit, so Model Vault is a later-stage choice.
Data handling: Cohere's FAQ says customer data on its platform is separated by logical segmentation, and in a private or cloud deployment through a partner your data is not shared with Cohere. It does not say whether platform data is used for training. A SOC 2 Type II report is available on request through Cohere's Trust Center.
Teams that fit Cohere
- Teams building search over their own documents. Embed 5 and Rerank 5 are the clearest reason to pick Cohere, and their prices are published.
- Products that serve many languages. Embed 5 covers more than 100 languages and Command A+ lists 48.
- Companies whose customers ask where data is processed. Private deployment and the Trust Center are on offer.
When Cohere is the wrong pick
- Anyone who needs a published, current price for a frontier chat model today. Cohere's pricing page does not give one.
- Teams that want to ship on the trial key. It bars commercial use and caps you at 1,000 calls a month.
- Founders counting on the $2,000 as a given. Cohere's page does not confirm the amount.
How Cohere compares with other credit offers
| Provider | What its official pages say | Our listing |
|---|---|---|
| Cohere | Trial key free, 1,000 calls a month, no commercial use. Embed 5 from $0.08 per 1M tokens | $2,000 in API credits |
| Together AI | Accelerator credits up to $15K, $30K or $50K by funding raised. Serverless gpt-oss-120B at $0.15 in and $0.60 out per 1M | $1,000 in credits |
| Anthropic | Startup page says applications were closed when we checked. Claude Haiku 5.5 at $0.10 in and $0.50 out per 1M for prompts up to 100,000 tokens | $2,000 in API credits |
Together AI's pricing page showed no matching models for embeddings or rerank when we checked, so for embeddings Cohere is the one with published prices in this table. For plain generation, Together and Anthropic both publish current per-token prices for their newer models, and Cohere's page does not.
If you want OpenAI credits too, see our guide to OpenAI API credits.
How to get the Cohere offer
- Open Cohere's startup program blog post at the link in our listing: cohere.com/blog/cohere-launches-startup-program. It is a Cohere page, not a SaaSOffers link.
- Create a Cohere account. A Trial key appears in the dashboard under API Keys.
- Contact Cohere through sales@cohere.com or its contact form and ask whether a startup credit exists, how much it is and when it expires. The blog post names no amount.
- If you get credits, ask whether they apply to Embed 5, Rerank 5 and Command A models, since the pricing page lists none of the current Command models.
- Move to a Production key through Billing and Usage before launch. Trial keys cannot serve customers.
Cohere setup tips
- Start with Embed 5 Fast. It costs 33% less than Pro per token ($0.08 against $0.12), and Fast and Pro share an embedding space, so you can move up later without a rebuild.
- Read the billed_units field in each API response, since Cohere says billed tokens can differ from total tokens.
- Documents over 500 tokens, query included, split into extra billed documents, so trim long passages before reranking.
Who Is This Deal For?
Early-Stage Startups
Seed and pre-seed companies looking to move fast without overspending on tools.
Growing SaaS Teams
Series A+ companies scaling their stack and optimizing software costs.
Solo Founders
Indie hackers and bootstrapped founders who need enterprise tools at startup prices.
Get $2,000 in API credits off Cohere
Premium deal. Upgrade once, unlock everything.
!Eligibility Requirements
AI startup building NLP features
Frequently Asked Questions
Everything you need to know about this startup deal.
Get the weekly deals digest
New verified startup deals every week. No spam, ever. Unsubscribe anytime.
Cohere guides
- Tools Every YC Batch Founder Should Claim First (2026)The exact tools every Y Combinator batch founder should claim in their first week, from infrastructure to CRM to legal. The complete YC starter stack for 2026.
- The Complete 2026 Tech Stack for AI StartupsThe complete tech stack for AI startups in 2026. LLM APIs, vector databases, infrastructure, monitoring, and the verified startup deals worth claiming for each layer.
Related Offers
Replicate
Used by 1,772 members
$500 in credits
Run open-source ML models in the cloud, deploy Llama, Stable Diffusion, and custom models via API without GPU management.
View offerConsensus
Used by 1,544 members
Free Plan
AI-powered research platform that finds, summarizes, and synthesizes scientific papers.
View offerLaxis
Used by 2,176 members
Free Plan
AI meeting assistant that records, transcribes, and generates actionable meeting notes.
View offerZoomInfo
Used by 2,014 members
Free Trial
B2B data platform for sales and marketing intelligence with contact databases.
View offerDeal Summary
Looking for more startup deals?
Browse all offers