Llama 3.3 70B Instruct
Open-weights chat model from Meta. One API key reaches all 10 providers, with automatic failover.
No card required. Start with 1 million free tokens.
About Llama 3.3 70B Instruct
Llama 3.3 70B Instruct is an open-weights chat model published by Meta, with a 128,000-token context window. Through TrustedRouter it runs on 10 providers, with options in the US and EU, from $0.14 per million input tokens. Three providers offer zero data retention, and Tinfoil also runs it in a confidential enclave.
- Model ID
meta-llama/llama-3.3-70b-instruct- Publisher
- Meta
- Modes
- chat, decide
- Billing
- Credits or your own provider key
Set it up with your agent
Paste this into Claude Code, Cursor or any coding agent.
Use Llama 3.3 70B Instruct through TrustedRouter in this project.
1. Read https://trustedrouter.com/llms.txt
2. Point the OpenAI-compatible client at
https://api.trustedrouter.com/v1 and read the key from
TRUSTEDROUTER_API_KEY.
3. Set the model to meta-llama/llama-3.3-70b-instruct.
4. Run the tests, then show me the diff before committing.import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.trustedrouter.com/v1",
api_key=os.environ["TRUSTEDROUTER_API_KEY"],
)
response = client.chat.completions.create(
model="meta-llama/llama-3.3-70b-instruct",
messages=[{"role": "user", "content": "What is the capital of France?"}],
)
print(response.choices[0].message.content)curl https://api.trustedrouter.com/v1/chat/completions \
-H "Authorization: Bearer $TRUSTEDROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "meta-llama/llama-3.3-70b-instruct",
"messages": [{"role": "user", "content": "What is the capital of France?"}]
}'Providers
Your key goes to the cheapest healthy provider and fails over on its own. Open a row for retention, location and weight format, or see how we check privacy claims and the live speed leaderboard.
| Provider | Input | Cached | Output | Cost per 1M in + out | Details |
|---|---|---|---|---|---|
Novita AI
Cheapest
|
$0.142 | — | $0.422 | $0.56 | |
|
Retention Verification: privacy unknown. No provider-ZDR claim is tracked here. Novita's privacy policy says personal information is not used for model training; customer-content processing is governed by customer agreements. Location Operator country: US Headquarters / registered office: San Francisco, California, United States. Company location source (reviewed 2026-09-27). Inference: Shared inference API; processing countries unconfirmed, no fixed-region commitment. No exhaustive inference-country list verified for the shared model API. Novita publishes worldwide GPU rental regions, but that separate product does not establish shared serverless model placement. Region pinning through TR: Not supported by the current TrustedRouter integration. Reviewed 2026-10-28. Weight format Unknown Price and speed Input $0.142425/1M · Cached input Not published · Output $0.422/1M TTFT: insufficient data Speed: insufficient data Last sample: |
|||||
AkashML
|
$0.211 | $0.106 | $0.549 | $0.76 | |
|
Retention Verification: privacy unknown. No contractual zero-retention, confidential-compute, or E2EE claim is tracked for TrustedRouter's AkashML account. This route is Standard. Location Operator country: Not verified Headquarters / registered office: Not verified. Inference: Inference location not verified; no country-specific residency commitment. No inference-location declaration has been verified for this provider. Region pinning through TR: No verified provider-specific location pin documented here. Weight format Unknown Price and speed Input $0.211/1M · Cached input $0.1055/1M · Output $0.5486/1M TTFT: 1.22 s Speed: insufficient data Last sample: |
|||||
Parasail
Zero retention
|
$0.232 | $0.116 | $0.528 | $0.76 | |
|
Retention Zero data retention. Tracked as ZDR for serverless and dedicated inference. Parasail documents no storage or logging of submitted input on those service paths, retention only while generating and delivering output, and no training on input or output. Batch service is excluded from this claim; TrustedRouter does not route Parasail traffic through batch. Location Operator country: US Headquarters / registered office: Not verified. Inference: Inference location not verified; no country-specific residency commitment. No inference-location declaration has been verified for this provider. Region pinning through TR: No verified provider-specific location pin documented here. Weight format Unknown Price and speed Input $0.2321/1M · Cached input $0.11605/1M · Output $0.5275/1M TTFT: insufficient data Speed: insufficient data Last sample: |
|||||
Telnyx
Zero retention
|
$0.633 | $0.633 | $0.633 | $1.27 | |
|
Retention Zero data retention. Telnyx's published policy states zero data retention for Telnyx-hosted /v2/ai/openai/chat/completions and /v2/ai/openai/embeddings: content is processed in memory and discarded after the response; request metadata is retained. TrustedRouter currently publishes hosted chat routes and adapts Responses to chat completions. Stateful Responses, AI Assistants with conversation history, third-party models and /v2/ai/embeddings are outside this ZDR scope. OpenAI-compatible embeddings may run in any region. DPA 3.1 restricts training of models available to others, with exceptions for written opt-in or data de-identified under 3.5. This is policy-backed ZDR, not confidential compute or verified end-to-end attestation. Location Operator country: US Headquarters / registered office: Austin, Texas, United States. Company location source (reviewed 2026-09-27). Inference: Multi-region, dynamically routed among the model's available Telnyx regions; no fixed inference-region commitment on this TrustedRouter route. Provider-wide GPU region set. Model availability is narrower; see the catalog snapshot below. Germany in the storage-locality documentation does not establish Germany as the inference country. Region pinning through TR: Not supported by the current TrustedRouter integration: it uses api.telnyx.com without Telnyx's strict region controls. Do not send Telnyx-specific region/mode fields as a TrustedRouter pin. Reviewed 2026-09-26. Inference regions and strict pinning Weight format Unknown Price and speed Input $0.633/1M · Cached input $0.633/1M · Output $0.633/1M TTFT: 0.89 s Speed: insufficient data Last sample: |
|||||
Weights & Biases
|
$0.749 | — | $0.749 | $1.50 | |
|
Retention Verification: no verified privacy claim. W&B Serverless Inference runs on CoreWeave infrastructure. W&B does not publish a zero-retention, confidential-compute, or end-to-end-encryption commitment for this service, so these routes use the Standard privacy tier. TrustedRouter does not enable W&B Weave tracing for provider calls. Location Operator country: US Headquarters / registered office: Not verified. Inference: Inference location not verified; no country-specific residency commitment. No inference-location declaration has been verified for this provider. Region pinning through TR: No verified provider-specific location pin documented here. Weight format Unknown Price and speed Input $0.74905/1M · Cached input Not published · Output $0.74905/1M TTFT: insufficient data Speed: insufficient data |
|||||
IO Intelligence
|
$0.505 | $0.252 | $1.15 | $1.66 | |
|
Retention Verification: privacy unknown. In its September 28, 2026 marketplace submission, io.net states that inference prompts and completions are processed only in memory, are not stored or logged, and are not used for training. Content-free operational metadata is retained for billing, capacity planning and abuse prevention. This is a provider declaration: the reviewed public privacy policy does not establish inference-specific zero-retention terms, and the applicable DPA has not been reviewed. Routes remain Standard pending that evidence. Confidential compute is offered on request, but this integration has no verified confidential-compute or provider-side E2EE guarantee. Location Operator country: US Headquarters / registered office: West Hollywood, California, United States (operating address). Company location source (reviewed 2026-09-28). Inference: United States and Canada, provider-declared; no US-only or fixed-country guarantee on this route. io.net declares US and Canadian serving locations and data residency in its September 28, 2026 marketplace submission. This is provider-wide information, not an independently verified per-model or per-request location. Region pinning through TR: Not supported by the current TrustedRouter integration. Reviewed 2026-09-28. API documentation (not a location guarantee) Weight format Unknown Price and speed Input $0.504923/1M · Cached input $0.252462/1M · Output $1.154803/1M TTFT: insufficient data Speed: insufficient data |
|||||
SambaNova
|
$0.633 | — | $1.27 | $1.90 | |
|
Retention Verification: privacy unknown. No contractual zero-retention, confidential-compute, or E2EE claim is tracked for TrustedRouter's SambaNova account. This route is Standard. Location Operator country: Not verified Headquarters / registered office: Not verified. Inference: Inference location not verified; no country-specific residency commitment. No inference-location declaration has been verified for this provider. Region pinning through TR: No verified provider-specific location pin documented here. Weight format Unknown Price and speed Input $0.633/1M · Cached input Not published · Output $1.266/1M TTFT: 1.07 s Speed: insufficient data Last sample: |
|||||
Databricks
|
$0.528 | — | $1.58 | $2.11 | |
|
Retention Verification: no verified privacy claim. Databricks Foundation Model APIs are a Designated Service and pay-per-token workloads are HIPAA compliant. Databricks may temporarily process or store inputs and outputs for abuse prevention, and may process data outside the originating cloud or region. This route is therefore standard privacy, not ZDR, confidential compute, or end-to-end encrypted. Location Operator country: US Headquarters / registered office: Not verified. Inference: Inference location not verified; no country-specific residency commitment. No inference-location declaration has been verified for this provider. Region pinning through TR: No verified provider-specific location pin documented here. Weight format Unknown Price and speed Input $0.5275/1M · Cached input Not published · Output $1.5825/1M TTFT: insufficient data Speed: insufficient data Last sample: |
|||||
Scaleway
|
$1.11 | — | $1.11 | $2.22 | |
|
Retention Verification: privacy unknown. Scaleway routes are discovered from its authenticated Model as a Service catalog and priced from Scaleway's first-party EUR price feed with a bounded FX reserve. No contractual ZDR, confidential-compute, or E2EE claim is tracked for this account, so these routes are Standard. Location Operator country: FR Headquarters / registered office: Not verified. Inference: Serverless: currently Paris, France, provider-declared; future expansion within Europe possible. Scaleway's FAQ places its current Serverless inference fleet in Paris. Dedicated deployments are a separate product. Region pinning through TR: Not supported by the current TrustedRouter integration. Reviewed 2026-09-27. Weight format Unknown Price and speed Input $1.109335/1M · Cached input Not published · Output $1.109335/1M TTFT: insufficient data Speed: insufficient data Last sample: |
|||||
Tinfoil
ConfidentialZero retention
|
$1.85 | — | $2.90 | $4.75 | |
|
Retention Confidential compute with end-to-end encryption and zero data retention. Tracked as a confidential inference provider with attested provider compute and no prompt/output logging claims. Location Operator country: US Headquarters / registered office: Not verified. Inference: Inference location not verified; no country-specific residency commitment. No inference-location declaration has been verified for this provider. Region pinning through TR: No verified provider-specific location pin documented here. Weight format FP8. Published serving configuration, reviewed 2026-09-27. Per-request precision is not verified. FP8 weight groups with dynamic FP8 activations; excluded layers retain higher precision. KV cache: unspecified. Price and speed Input $1.84625/1M · Cached input Not published · Output $2.90125/1M TTFT: insufficient data Speed: insufficient data Last sample: |
|||||
Append :nitro to the model ID, or send "provider": {"sort": "throughput"}, to route to the fastest provider instead.
Questions
What model ID should I use for Llama 3.3 70B Instruct?
Use meta-llama/llama-3.3-70b-instruct as the model field with the TrustedRouter OpenAI-compatible API. Credits routing and provider fallback happen behind that model ID.
Do you log my prompts or outputs?
No. Prompt and output content are intentionally excluded from the control plane and metadata systems. Optional content export requires explicit destination-level configuration.
Which providers serve Llama 3.3 70B Instruct?
TrustedRouter currently lists Tinfoil, Parasail, Telnyx, Novita AI, Databricks, AkashML, SambaNova, IO Intelligence, Scaleway, and Weights & Biases for Llama 3.3 70B Instruct. Provider availability and routing eligibility can change as catalog and health data update.
How much does Llama 3.3 70B Instruct cost through TrustedRouter?
The current lowest Credits input price is $0.142425/1M and the lowest Credits output price is $0.422/1M. Prices are per one million tokens and come from the current route catalog.
How do I require zero data retention for Llama 3.3 70B Instruct?
Set provider.min_privacy to zdr on the request. TrustedRouter considers only routes with a recorded zero-data-retention posture and fails closed if no eligible route remains.
Llama 3.3 70B Instruct vs other models
Start building with Llama 3.3 70B Instruct.
No card required. Start with 1 million free tokens.