Models / Uncensored / failspy
Llama 3 70B Abliterated v3.5 API pricing
$2.10 in, $2.10 out per million tokens.
Price · per million tokens, USD
| Tokens | failspy list price | UplinkSwitch | Saving |
|---|---|---|---|
| Input | — | $2.10 | — |
| Cached input | Billed at the input rate. | ||
| Output | — | $2.10 | — |
Llama 3 70B Abliterated v3.5 specs
- Context window
- 8,192 tokens
- Input
- text
Llama 3 70B Abliterated v3.5: failspy’s orthogonalized Llama · from the author of the abliteration cookbook
Llama 3 70B Abliterated v3.5 is Meta's Llama 3 70B Instruct with orthogonalized weights by failspy, who published a cookbook for the method (its card). The method comes from the paper “Refusal in LLMs is mediated by a single direction”. In v3.5 only one layer is modified, which the card says removed the moralizing disclaimers; it also fixes the tokenizer. It is served as llama-3-70b-abliterated at $2.10 input and $2.10 output per million tokens, one of 49 builds in the uncensored LLM API, on the same key as every other model. Added 2026-10-20.
| Llama 3 70B Abliterated v3.5 | |
|---|---|
| Base model | Llama 3 70B Instruct |
| Made by | failspy |
| Method | Orthogonalized weights, one layer changed |
| License on the card | llama3 |
| Context window here | 8,192 tokens |
| Downloads, 30 days to 2026-10-20 | 8,715 |
| Card | failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5 |
From the Hugging Face card and its API record, read 2026-10-20.
curl https://api.uplinkswitch.com/v1/chat/completions \
-H "authorization: Bearer sk-uplink-live-..." \
-H "content-type: application/json" \
-d '{"model":"llama-3-70b-abliterated","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'Llama 3 70B Abliterated v3.5 next to the other uncensored models · per million tokens, live
A chat turn is 2,000 input and 500 output tokens. Live from the catalog, ordered by the cost of 1,000 turns.
Worked for Llama 3 70B Abliterated v3.5: 1,000 turns are 2,000,000 input tokens and 500,000 output tokens. At $2.10 input and $2.10 output per million tokens, that is $4.20 + $1.05 = $5.25, paid from prepaid credit with no subscription.
Llama 3 70B Abliterated v3.5 API · base URL and key
from openai import OpenAI
client = OpenAI(
base_url="https://api.uplinkswitch.com/v1",
api_key="sk-uplink-live-...",
)
completion = client.chat.completions.create(
model="llama-3-70b-abliterated",
messages=[{"role": "user", "content": "Summarize this in one line."}],
)
print(completion.choices[0].message.content)import OpenAI from 'openai';
const client = new OpenAI({
baseURL: 'https://api.uplinkswitch.com/v1',
apiKey: 'sk-uplink-live-...',
});
const completion = await client.chat.completions.create({
model: 'llama-3-70b-abliterated',
messages: [{ role: 'user', content: 'Summarize this in one line.' }],
});
console.log(completion.choices[0].message.content);curl https://api.uplinkswitch.com/v1/chat/completions \
-H "authorization: Bearer sk-uplink-live-..." \
-H "content-type: application/json" \
-d '{"model":"llama-3-70b-abliterated","messages":[{"role":"user","content":"Hello"}]}'What would it cost? · paste a prompt, priced before you send it
Input is counted at about 4 characters per token.
Cheapest first. A real request is billed on the provider's own token count. Same estimate as JSON: POST https://api.uplinkswitch.com/v1/estimate, no key needed.
Llama 3 70B Abliterated v3.5 questions
- What is Llama 3 70B Abliterated v3.5?
- Llama 3 70B Abliterated v3.5 is a model from failspy, with a 8,192-token context window, taking text input. Through UplinkSwitch it answers the OpenAI SDK on one key as llama-3-70b-abliterated.
- How much does the Llama 3 70B Abliterated v3.5 API cost?
- $2.10 per million input tokens and $2.10 per million output tokens.
- Is Llama 3 70B Abliterated v3.5 free?
- It is pay as you go: prepaid credit from $10, with no subscription or monthly fee, billed per token. A 1,000-token prompt with a 500-token reply costs about $0.0032.
- What is the Llama 3 70B Abliterated v3.5 context window?
- Llama 3 70B Abliterated v3.5 has a 8,192-token context window. It takes text input.
- What is the Llama 3 70B Abliterated v3.5 model id?
- llama-3-70b-abliterated. Send it as the model through the OpenAI SDK with base URL https://api.uplinkswitch.com/v1.
- How do I get a Llama 3 70B Abliterated v3.5 API key?
- Create one in the console at uplinkswitch.ai/console: prepaid credit from $10, no subscription, and the key is live at once. The same key calls Llama 3 70B Abliterated v3.5 and every other model in the catalog; send it as the bearer token with base URL https://api.uplinkswitch.com/v1.
- Is there a Llama 3 70B Abliterated v3.5 API?
- Yes: llama-3-70b-abliterated on the OpenAI format at https://api.uplinkswitch.com/v1, $2.10 input and $2.10 output per million tokens. No weights to download and no GPU of your own; the key also runs every other model in the catalog.
- What is Llama 3 70B Abliterated v3.5 based on?
- Meta’s Meta-Llama-3-70B-Instruct, with its weights orthogonalized against the refusal direction by failspy; v3.5 changes a single layer, per its card.
- What license does Llama 3 70B Abliterated v3.5 have?
- The card’s license tag is llama3, Meta’s Llama 3 license.
- What does Llama 3 70B Abliterated v3.5 cost per 100 messages?
- $0.525 for 100 chat turns of 2,000 input and 500 output tokens each, at $2.10 input and $2.10 output per million tokens, from prepaid credit with no subscription.
- Llama 3 70B Abliterated v3.5 or Llama 3.3 70B Abliterated?
- Both are abliterated Llama 70B Instruct models: v3.5 is failspy’s Llama 3, the other huihui-ai’s Llama 3.3 with the longer window. Llama 3 70B Abliterated v3.5 is $2.10 / $2.10 per million tokens in / out, $5.25 for 1,000 chat turns; Llama 3.3 70B Abliterated is $2.10 / $2.10 per million tokens in / out, $5.25 for 1,000 chat turns. Both run on the same key, so send the same prompt to each.