Together AI

together.ai · Compute, Inference, Storage · 5 records · JSON

CodeSandbox SDK Computetrialemailno cardDretention-clamped
Together Code Sandbox Computepaidemail?card?Drequest-rate
Together AI Computepaidemail?card?Drequest-rate
Together AI InferencepaidcardcardD
Together AI Files API (with 100 GB fine-tune objects) Storagefails: identitycardcardR

CodeSandbox SDK Compute

trialemailno cardDretention-clamped

docs.together.ai/docs/together-code-sandbox-legacy · container and VM sandbox · compute/codesandbox-sdk · JSON · compare

Free tier	the rate card it carries is labelled legacy unit mixed; metered on wall-clock; Together recommends its successor for new integrations
Price	none stated
Per-object cap	not applicable
Retention	hibernation stops compute billing; the disk term was not read
Egress	not published on the page read
Identity	email an account; the tier gate was not read
Limits	request-rate Build allows 10 concurrent VMs retention-clamped the published rates are labelled legacy source 10, 20 and 40 VM credits an hour at USD 0.01486 each on the legacy card
Binding	retention-clamped the rate card is labelled legacy and Together points at a successor, so the product's own future ends the offer before any quota does
Interface	unknown an SDK that creates VMs from credits
Tooling	the CodeSandbox SDK, now legacy under Together
Composes with	none for new work: the rates are legacy and the product has a successor
Reliability	no reliability event recorded
Notes	Recorded so a reader does not build on a legacy rate card. The credit is real and the product is being replaced.
Evidence rung	documented re-read live on 2026-10-22 for the free-tier and limit figures
Checked	2026-10-23 (live); stale on 2026-11-22

Evidence (2)

date	source	what it proved
2026-10-02	docs.together.ai/docs/together-code-sandbox-legacy	the rate card is labelled legacy and Together recommends its successor for new integrations
2026-10-22	docs.together.ai/docs/together-code-sandbox-legacy	read live: the rate card is deprecated and a successor SDK is named

More fields (9)

{
 "shape": null,
 "billing_model": "allocated",
 "session_cap": null,
 "keep_rate": null,
 "persistent_storage": null,
 "network_out": null,
 "free_tier_kind": "one-time-credit",
 "granularity": {
  "min_billed_s": 60,
  "rounding_s": null
 },
 "burst_x": 1.0
}

Unknown or withheld values (6)

$.egress.metered egress-unknown
$.interface interface-unknown
$.limits[0].source withheld
$.per_object_cap.notes process
$.per_object_cap.stated not-applicable
$.reliability_note none-observed

Together Code Sandbox Compute

paidemail?card?Drequest-rate

www.together.ai/pricing · container and VM sandbox · compute/together-code-sandbox · JSON · compare

Free tier	no separate sandbox allowance: model inference bills per million tokens from $0.00 on Ternary Bonsai to $15.00 an output MTok, and a code sandbox draws on the same account unit none; metered on none
Price	none stated
Per-object cap	not stated
Retention	not stated
Egress	not stated
Identity	email (placeholder) not published
Limits	request-rate per million input and output tokens, from $0.00 on Ternary Bonsai to $15.00 an output MTok source per-object-cap dedicated capacity is reserved in throughput units, where tokens a minute depends on the model source not stated
Binding	request-rate
Interface	unknown
Tooling	not stated
Composes with	not stated
Reliability	no reliability event recorded
Evidence rung	documented the pricing page was read on 2026-10-22 and carries model and audio rates. It publishes no sandbox-specific free allowance, so the record's free-tier claim is not established and the verdict moves to paid-only
Checked	2026-10-23 (live); stale on 2026-11-22

Evidence (2)

date	source	what it proved
2026-10-02	www.together.ai/pricing	not published
2026-10-22	www.together.ai/pricing	the pricing page was read on 2026-10-22 and is a per-million-token model table with 'Kimi K3 $3.00 input, $15.00 output', 'GLM-5.3 $1.40 / $4.40', 'Ternary Bonsai 27B $0.00', and 'Price per audio minute'; the sandbox product is billed through the same meters

More fields (17)

{
 "talaria_m10h_usd": 44.64,
 "talaria_m24h_usd": 107.14,
 "nemo_m10h_usd": 9.0,
 "keep_rate": 1.0,
 "keep_rate_basis": null,
 "billing_model": "unclassified",
 "shape": null,
 "session_cap": {
  "wall_clock_min": null,
  "idle_reap_s": null,
  "restart_required": false,
  "note": "a sandbox draws on the same token and audio meters; the page states no sandbox clock"
 },
 "persistent_storage": null,
 "network_out": null,
 "price_note": null,
 "free_tier_kind": "none",
 "granularity": {
  "min_billed_s": 60,
  "rounding_s": null
 },
 "burst_x": 2.0,
 "burst_bill": "min 1 h",
 "ps": null,
 "net": {
  "out": true,
  "in": true,
  "ports": null,
  "static_ip": null,
  "ipv6": null,
  "allowance": null
 }
}

Unknown or withheld values (17)

$.chaining not-stated
$.egress.metered egress-unknown
$.egress.terms not-stated
$.evidence[0].proved process
$.extras.keep_rate_basis process
$.extras.price_note process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits_note not-stated
$.per_object_cap.notes process
$.per_object_cap.stated not-stated
$.reliability_note none-observed
$.retention.notes not-extracted
$.retention.stated not-stated
$.tooling not-stated

Together AI Compute

paidemail?card?Drequest-rate

www.together.ai/pricing · GPU cloud · compute/together-gpu · JSON · compare

Free tier	no free GPU allowance: GPU capacity is priced per GPU-hour on demand or reserved, and inference bills per million tokens from $0.00 on Ternary Bonsai to $15.00 an output MTok unit none; metered on none
Price	none stated
Per-object cap	not stated
Retention	not stated
Egress	not stated
Identity	email (placeholder) not published
Limits	request-rate GPU capacity priced per GPU per hour, on demand or reserved; reserved capacity is sold in throughput units source request-rate inference bills per million input and output tokens, from $0.00 to $15.00 an output MTok source excluded as gpu-only
Binding	request-rate
Interface	unknown
Tooling	not stated
Composes with	not stated
Reliability	no reliability event recorded
Evidence rung	documented the pricing page was read on 2026-10-22 and its GPU section prices per GPU per hour. It publishes no free GPU allowance, so the record's free-tier claim is not established and the verdict moves to paid-only
Checked	2026-10-23 (live); stale on 2026-11-22

Evidence (2)

date	source	what it proved
2026-10-02	www.together.ai/pricing	not published
2026-10-22	www.together.ai/pricing	the pricing page was read on 2026-10-22 and its GPU section states 'All prices are per GPU per hour' with on-demand and reserved columns, alongside per-million-token model rates from 'Ternary Bonsai 27B $0.00' to 'Kimi K3 $3.00 / $15.00'

More fields (14)

{
 "talaria_m10h_usd": null,
 "talaria_m24h_usd": null,
 "nemo_m10h_usd": null,
 "keep_rate": null,
 "keep_rate_basis": "no field in the source",
 "billing_model": "unclassified",
 "shape": {
  "vcpu": null,
  "ram": null,
  "disk": null,
  "gpu_class": null,
  "gpu_count": null,
  "note": "priced per GPU per hour with on-demand and reserved rates"
 },
 "session_cap": {
  "wall_clock_min": null,
  "idle_reap_s": null,
  "restart_required": false,
  "note": "a GPU instance bills per hour; the page states no reap"
 },
 "persistent_storage": null,
 "network_out": null,
 "price_note": null,
 "free_tier_kind": "none",
 "ps": {
  "gib": null,
  "price": "managed storage is a separate section",
  "durability": "managed storage",
  "on_stop": true,
  "on_destroy": false,
  "retention": "while the volume exists"
 },
 "net": null
}

Unknown or withheld values (15)

$.chaining not-stated
$.egress.metered egress-unknown
$.egress.terms not-stated
$.evidence[0].proved process
$.extras.price_note process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.per_object_cap.notes process
$.per_object_cap.stated not-stated
$.reliability_note none-observed
$.retention.notes not-extracted
$.retention.stated not-stated
$.tooling not-stated

Together AI Inference

paidcardcardD

docs.together.ai/docs/billing · inference endpoint · inference/together · JSON · compare

Free tier	No free trial: 'Together AI does not currently offer free trials. Access to the Together platform requires a minimum $15 credit purchase.' unit none; metered on none
Price	none stated
Per-object cap	not applicable: an inference endpoint stores nothing for the caller
Retention	not applicable
Egress	the response is the payload; no separate egress term is published
Identity	card an account and an API key; a card at signup
Binding	none
Interface	rest
Tooling	any OpenAI-compatible client
Composes with	no free lane: the platform requires a minimum USD 15 purchase, so there is no trial credit to pair with an agent
Reliability	no reliability event recorded; no key was available
Notes	The compute topic also carries this provider under separate slugs. The vendor denies a free trial in its own billing documentation, so the record is paid-only.
Evidence rung	documented the billing documentation was read on 2026-10-23 and states that Together offers no free trial and requires a USD 15 minimum purchase. The unauthenticated 401 probe from 2026-10-22 establishes only that the endpoint exists and gates on a key
Checked	2026-10-24 (live); stale on 2026-11-23

Evidence (3)

date	source	what it proved
2026-10-22	probe:https://api.together.xyz/v1/chat/completions	an unauthenticated POST returned 401 with a 372-byte body, {"error":{"message":"Missing API key..."}}. The endpoint exists and gates on a key; no key existed here, so no quota was read
2026-10-22	docs.together.ai/docs/rate-limits	the rate-limits page returned 200 with 388051 bytes and describes dynamic per-model limits and throughput guarantees; no free-tier allowance was found in the served text
2026-10-23	docs.together.ai/docs/billing	'Together AI does not currently offer free trials. Access to the Together platform requires a minimum $15 credit purchase.', quoted

Unknown or withheld values (1)

$.retention.stated not-applicable

Corrections

2026-10-23 The credit ladder was re-read, and eight claims moved

Together AI Files API (with 100 GB fine-tune objects) Storage

fails: identitycardcardR

docs.together.ai/docs/fine-tuning/data-preparation · LLM provider /v1/files · storage/together-ai-files-api · JSON · compare

Free tier	100 GB per training file (JSONL or Parquet); SDK constant MAX_FILE_SIZE_GB = 50.1 and MULTIPART_THRESHOLD_GB = 5.0. No published total-storage quota. unit GB; metered on stored-bytes
Price	$163.84 (Managed Storage at $0.16/GiB/mo) — 40-80x the cheap-paid floor; egress free
Per-object cap	per-object-cap 100 GB (100000000000 bytes)
Retention	not stated
Egress	No egress fee on the Files API. Their separate Managed Storage product is explicitly advertised with 'zero egress fees'.
Identity	card card
Limits	per-object-cap per-object-cap 100 GB source per-object-cap 100 GB. Multipart kicks in above 5 GB; per-part upload timeout 300 s. A client-side check (`check_file`) validates JSONL structure before upload, which is a soft write gate for the fine-tune purpose.
Binding	none the quota is real and unreachable: identity is the binding constraint, and it is not one of the six limits kinds
Interface	unknown POST https://api.together.xyz/v1/files/upload (and /v1/files/{id}/content); OpenAI-ish, `purpose` in {fine-tune, calibration, eval, batch-api}
Tooling	together-python (`tg files upload`), curl multipart
Composes with	100 GB single objects are the largest per-object cap found in the entire LLM-provider lane — 200x OpenAI's 512 MB. If a total quota is genuinely absent, this is the highest-leverage per-object number in this dossier. Unverified.
Reliability	Together has no free tier — 'no free trials', $5 minimum prepaid credit required. That already fails the client's constraint. Read path unverified: /v1/files/{id}/content exists but every probe returns 401. Managed Storage is $0.16/GiB/month = $163.84/TB/month, which is 40-80x the cheap-paid lane and therefore useless even though egress is free.
Notes	100 GB per training file (JSONL or Parquet); SDK constant MAX_FILE_SIZE_GB = 50.1 and MULTIPART_THRESHOLD_GB = 5.0. No published total-storage quota.
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers	`$.limits[0].value` limit:per-object-cap 100.0 GB `$.price_per_tb_month` price-per-tb USD0.16 GiB / month

Evidence (6)

date	source	what it proved
2026-10-21	docs.together.ai/docs/fine-tuning/data-preparation	'The file size limit for either format is 100 GB.'
2026-10-21	github.com/togethercomputer/together-python/blob/main/src/to	'MAX_FILE_SIZE_GB = 50.1', 'MULTIPART_THRESHOLD_GB = 5.0', 'MULTIPART_UPLOAD_TIMEOUT = 300'
2026-10-21	www.together.ai/managed-storage	'Object storage and parallel filesystems optimized for AI, provisioned and fully managed … with zero egress fees'
2026-10-21	markaicode.com/pricing/together-ai-pricing/	'storage ($0.16/GiB/month for the parallel filesystem used by GPU Clusters)'
2026-10-21	www.flowtools.co/alternatives/openrouter	'Together AI does not currently offer free trials … minimum $5 purchase'
2026-10-21	api.together.xyz/v1/files	PROBE: curl -D - -> 401, Content-Length 148, {"error":{"message":"No API key provided…"}}; /v1/files/abc/content -> 401/148

More fields (2)

{
 "dead_end": {
  "failure_class": "quota",
  "why": "100 GB per training file (JSONL or Parquet); SDK constant MAX_FILE_SIZE_GB = 50.1 and MULTIPART_THRESHOLD_GB = 5.0. No published total-storage quota.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 },
 "price_verbatim": "$163.84 (Managed Storage at $0.16/GiB/mo) — 40-80x the cheap-paid floor; egress free"
}

Unknown or withheld values (6)

$.extras.dead_end.retry_when process
$.free_tier.notes process
$.interface interface-unknown
$.per_object_cap.notes not-extracted
$.retention.notes not-extracted
$.retention.stated not-extracted