Together AI serverless inference token pricing (per 1M tokens, 2026-10-04 list price)

USD per 1M TOKENS, LIST PRICE, observed 2026-10-04. Qwen3.5-397B-A17B: input $0.60 ($0.35 cached) / output $3.60. Qwen3.8-2.4T-A95B: $2.00 input ($0.25 cached) / $6.00 output. Qwen3.7-Max $1.50 ($0.30 cached) / $4.50. Qwen3.7-Plus $0.32 / $1.28. Qwen3.6-Plus $0.50 / $3.00. Qwen3.8 Flash $0.09 / $0.28. Qwen3.5 9B $0.17 / $0.25. Llama 3.3 70B $1.04 / $1.04 - identical input and output, an unusual shape worth noting. Llama 3 8B Instruct Lite $0.14 / $0.14. Qwen2.5 7B Instruct Turbo $0.30 / $0.30. Throughput-tier FP8 variant: Qwen3 235B A22B Instruct 2507 FP8 Throughput $0.20 / $0.60. price_usd = Qwen3.5-397B-A17B input PERISHABLE-FIGURE STAMP (added 2026-10-04, uniform across data/supply/ so it is greppable). PRICE CLASS: PUBLISHED LIST PRICE - an asking rate a new customer is quoted, NOT a realised, discounted or negotiated transaction rate. A real bill is at or below this and typically well below it once commitments, credits and committed-use discounts apply. OBSERVED: 2026-10-04 RE-VERIFY BY: 2027-01-04 - the repo's quarterly re-verification bar. After that date this figure is presumed drifted and must be re-checked against its source before it is quoted again. A price without its observation date is noise; this stamp exists so a reader can tell a fresh figure from a stale one without reading prose. UNIT: USD per 1M TOKENS, LIST PRICE, observed 2026-10-04 READING RULE: No further perishable flag is published for this figure.

lead_time_weeks is null because NO NUMBER APPLIES, and the reason is specific rather than generic: this record prices a metered service term (tokens, egress, storage) rather than hardware capacity, so there is no order-to-delivery interval to report at all. Calling this 'no published lead time' would be wrong in the other direction - nothing was withheld.

no EAR citation: no controlled item is transferred by this transaction; not an EAR citation: provider service terms and end-use policy (the AI Diffusion Rule briefly tiered service access before its 2025-05-13 rescission)

MECHANISM: a token-priced API, so no accelerator crosses a border and the 3A090 hardware licence regime does not attach to the transaction. Export control becomes an END-USER AND TERMS question instead: the provider applies geographic and end-use restrictions to service access, and a customer can still be refused under sanctions or end-use policy wherever the compute sits. A reader targeting a restricted destination should read the provider's terms, not the EAR.

Together's line item that matters most for inference economics is the SEPARATE GPU Clusters product, priced per GPU-hour on the same page (recorded separately as rent-together-gpu-clusters-b200-hourly in this slice). The serverless table above is per-token and sits on different hardware economics; do not derive one from the other. Two traps: (a) the cached-input rate is not a fixed discount - it is $0.25 against $2.00 (8x) for Qwen3.8-2.4T-A95B but $0.35 against $0.60 (1.7x) for Qwen3.5-397B, so 'use the cache' saves wildly different amounts per model; (b) several Llama rows are priced symmetrically input=output, which makes an input-heavy workload cost materially more on Together than the headline output rate suggests. No regional GPU list and no GPU hourly rate in the serverless section. EXPORT CONTROL (2026-10-04): NOT APPLICABLE IN THE HARDWARE-SHIPMENT SENSE - this record prices a token-based API, so no accelerator is transferred across a border and the ECCN 3A090 hardware licence regime does not attach to the transaction. Recorded null rather than false precisely because a token API is where export control instead becomes an END-USER/TERMS question: the provider applies geographic and end-user restrictions to service access (the AI Diffusion Rule briefly put a tiered country structure on exactly this kind of service before being rescinded on 2025-05-13), and a customer can still be refused under sanctions or end-use policy regardless of where the compute sits. Anyone reading this record for a restricted destination should look at the provider's terms, not at the EAR. Sources for the regime: spl-bis-2025-05-ai-diffusion-rescission, spl-bis-2026-05-d5-guidance. EXPORT CONTROL (2026-10-04): NOT APPLICABLE IN THE HARDWARE-SHIPMENT SENSE - this record prices a token-based API, so no accelerator crosses a border and the ECCN 3A090 hardware licence regime does not attach to the transaction. Recorded NULL rather than false precisely because a token API is where export control instead becomes an END-USER and TERMS question: the provider applies geographic and end-user restrictions to service access (the AI Diffusion Rule briefly put a tiered country structure on exactly this kind of service before being rescinded and unenforced from 2025-05-13), and a customer can still be refused under sanctions or end-use policy regardless of where the compute sits. For a restricted destination, read the provider's terms, not the EAR. Sources: spl-bis-2025-05-ai-diffusion-rescission, spl-bis-2026-05-d5-guidance. LEAD TIME: null - correct and not a gap. A token-priced serverless API is provisioned on request, so there is no lead time to publish; the binding constraint when a model is at capacity is provider-side queuing and any per-account concurrency limit, which this repo records in notes where a vendor publishes one (see the DeepSeek concurrency figures already here). Deliberately NOT recorded: any 'model capacity lead time' estimate, because vendors publish none and the SEO 'GPU lead time' numbers that circulate do not apply to a rented endpoint. Probe log in source spl-negative-no-oem-publishes-lead-time. ACCELERATOR LINK ADDED 2026-10-04: the serverless table is served on rented NVIDIA capacity (B200-class in this slice; the same page's separate GPU Clusters product is the per-GPU-hour counterpart, recorded in ppl-sup-together-* and together-gpu-clusters-h100). Linked to B200 to make the warning in the notes concrete: per-token serverless rates and per-GPU-hour cluster rates in this vendor's slice sit on different hardware economics and must not be derived from one another.