Skip to content
Model infoModelsSnapshot: 15 Oct 2026

ByteDance / Seed

Seed 2.1 Turbo (260628)

Seed 2.1 Turbo language-model revision 260628, with documented reasoning and tool calling through ModelArk.

Language models (LLMs)
availableClosed weightsText generation
Official documentation (opens in a new tab)Reviewed 15 Oct 2026
14 of 20 specification fields collected
Input token limit256KSee specification notes below
Maximum output256KSee specification notes below
Input / 1M tokens$0.50BytePlus ModelArk
Output / 1M tokens$2.50USD · observed 12 Oct 2026

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
256KSource for Context / input limit (opens in a new tab)

Provider K-token notation retained. Output limit includes chain-of-thought tokens.

Maximum output
256KSource for Maximum output (opens in a new tab)

Provider K-token notation retained. Output limit includes chain-of-thought tokens.

Input modalities
Text, Image, VideoSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)
Reasoning
SupportedSource for Reasoning (opens in a new tab)
Tool calling
SupportedSource for Tool calling (opens in a new tab)
Structured output
SupportedSource for Structured output (opens in a new tab)

The catalog labels structured output beta. JSON schema is explicitly recommended for the April 2026 Mini and Lite releases.

Streaming
SupportedSource for Streaming (opens in a new tab)

This exact model ID occurs in the streaming guide’s request or response examples.

Batch processing
SupportedSource for Batch processing (opens in a new tab)
Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
Not collected
Parameters
Not collected
Architecture
Not collected
Languages
Not collected
Context caching
SupportedSource for Context caching (opens in a new tab)
Maximum input tokens
256KSource for Maximum input tokens (opens in a new tab)

Provider K-token notation retained. Output limit includes chain-of-thought tokens.

Maximum thinking tokens
256KSource for Maximum thinking tokens (opens in a new tab)

Provider K-token notation retained. Output limit includes chain-of-thought tokens.

Published request quota
500 Max RPM; 1000K Max TPMSource for Published request quota (opens in a new tab)

Published quota, not a throughput or latency benchmark.

Flex inference
SupportedSource for Flex inference (opens in a new tab)

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

BytePlus ModelArk · standard

Provider source (opens in a new tab)
API model ID
dola-seed-2-1-turbo-260628
Access
ModelArk API access requires an eligible BytePlus account and enabled model. Published quotas are account-dependent.
Regions
Not collected
Service tier
standardSource for Service tier (opens in a new tab)
INPUT / 1M TOKENS$0.50
OUTPUT / 1M TOKENS$2.50

standard non-audio rate table. Full prompt length selects the rate for the entire request; thinking tokens count toward output. Provider lists the Turbo family price, linked to the sole Turbo revision in the reviewed catalog. Finite pricing bounds do not override modality-specific API input limits. Cached input price: $0.10 / 1M tokens (base tier).

Pricing by prompt length, USD / 1M tokens
Input lengthInputOutputCached input
Prompt length [0, 256] K tokens$0.50$2.50$0.10
USD · observed 12 Oct 2026 · Pricing evidence (opens in a new tab)
Published billing units · original currency · observed 12 Oct 2026
CategoryUsageRateUnit
CachingExplicit cache storage1$0.00831M tokens/hour

Rate conditions

  1. Prompt length [0, 256] K tokens. Explicit cache only; implicit storage is free. Partial natural hours billed as whole hours. Pricing-table units used; the cache guide uses a different generic unit description.
Pricing source (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

BytePlus ModelArk · flex

Provider source (opens in a new tab)
API model ID
dola-seed-2-1-turbo-260628
Access
ModelArk API access requires an eligible BytePlus account and enabled model. Published quotas are account-dependent.
Regions
Not collected
Service tier
flexSource for Service tier (opens in a new tab)
Flex inference
SupportedSource for Flex inference (opens in a new tab)
INPUT / 1M TOKENS$0.25
OUTPUT / 1M TOKENS$1.25

flex non-audio rate table. Full prompt length selects the rate for the entire request; thinking tokens count toward output. Provider lists the Turbo family price, linked to the sole Turbo revision in the reviewed catalog. Finite pricing bounds do not override modality-specific API input limits. Cached input price: $0.10 / 1M tokens (base tier).

Pricing by prompt length, USD / 1M tokens
Input lengthInputOutputCached input
Prompt length [0, 256] K tokens$0.25$1.25$0.10
USD · observed 12 Oct 2026 · Pricing evidence (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

BytePlus ModelArk · batch

Provider source (opens in a new tab)
API model ID
dola-seed-2-1-turbo-260628
Access
ModelArk API access requires an eligible BytePlus account and enabled model. Published quotas are account-dependent.
Regions
Not collected
Service tier
batchSource for Service tier (opens in a new tab)
Batch processing
SupportedSource for Batch processing (opens in a new tab)
INPUT / 1M TOKENS$0.25
OUTPUT / 1M TOKENS$1.25

batch non-audio rate table. Full prompt length selects the rate for the entire request; thinking tokens count toward output. Provider lists the Turbo family price, linked to the sole Turbo revision in the reviewed catalog. Finite pricing bounds do not override modality-specific API input limits. Cached input price: $0.10 / 1M tokens (base tier).

Pricing by prompt length, USD / 1M tokens
Input lengthInputOutputCached input
Prompt length [0, 256] K tokens$0.25$1.25$0.10
USD · observed 12 Oct 2026 · Pricing evidence (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Volcengine Ark (China) · Standard online; other family tariff modes are listed separately and require eligibility.

Provider source (opens in a new tab)
API model ID
doubao-seed-2-1-turbo-260628
Access
Native service activation and an eligible Volcengine account/API key required. No inference executed.
Regions
cn-beijing
Context / input limit
256KSource for Context / input limit (opens in a new tab)

Shared context; provider K notation retained.

Maximum input tokens
256KSource for Maximum input tokens (opens in a new tab)
Maximum output
256KSource for Maximum output (opens in a new tab)

Answer and reasoning still share the context budget.

Published chain-of-thought limit
256KSource for Published chain-of-thought limit (opens in a new tab)
Input modalities
Text, Image, VideoSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)
Reasoning
SupportedSource for Reasoning (opens in a new tab)
Tool calling
SupportedSource for Tool calling (opens in a new tab)
Structured output
SupportedSource for Structured output (opens in a new tab)
Published request quota
Max RPM: 500 Max TPM: 1000000Source for Published request quota (opens in a new tab)

Published quota, not a guaranteed benchmark.

Service tier
Standard online; other family tariff modes are listed separately and require eligibility.Source for Service tier (opens in a new tab)
Streaming
Not collected

Native version-specific streaming contract not established by this catalog row.

Batch processing
Not collected

Family batch tariff is not proof of this dated native endpoint entitlement.

Fine-tuning
Not collected

International fine-tuning availability is not projected onto this native endpoint.

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Volcengine Standard online family tariffNon-audio input · Input length [0, 256] (thousand tokens)1CN¥3.001M tokens
Volcengine Standard online family tariffExplicit cache storage · Input length [0, 256] (thousand tokens)1CN¥0.0171M tokens/hour
Volcengine Standard online family tariffCached non-audio input · Input length [0, 256] (thousand tokens)1CN¥0.601M tokens
Volcengine Standard online family tariffOutput · Input length [0, 256] (thousand tokens)1CN¥15.001M tokens
Volcengine Low-latency online family tariffNon-audio input · Input length [0, 256] (thousand tokens)2CN¥6.001M tokens
Volcengine Low-latency online family tariffCached non-audio input · Input length [0, 256] (thousand tokens)2CN¥1.201M tokens
Volcengine Low-latency online family tariffOutput · Input length [0, 256] (thousand tokens)2CN¥30.001M tokens
Volcengine Flex online family tariffNon-audio input · Input length [0, 256] (thousand tokens)3CN¥1.501M tokens
Volcengine Flex online family tariffCached non-audio input · Input length [0, 256] (thousand tokens)3CN¥0.601M tokens
Volcengine Flex online family tariffOutput · Input length [0, 256] (thousand tokens)3CN¥7.501M tokens
Volcengine Batch inference family tariffNon-audio input · Input length [0, 256] (thousand tokens)4CN¥1.501M tokens
Volcengine Batch inference family tariffCached non-audio input · Input length [0, 256] (thousand tokens)4CN¥0.601M tokens
Volcengine Batch inference family tariffOutput · Input length [0, 256] (thousand tokens)4CN¥7.501M tokens

Rate conditions

  1. doubao-seed-2.1-turbo, Standard online; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
  2. doubao-seed-2.1-turbo, Low-latency online; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
  3. doubao-seed-2.1-turbo, Flex online; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
  4. doubao-seed-2.1-turbo, Batch inference; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
Pricing source (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

BytePlus ModelArk · standard

dola-seed-2-1-turbo-260628
Documented
Support evidence (opens in a new tab)
How it works
Implicit: Responses API, Chat API. Explicit: Responses API, Prefix caching, Session caching.Source for How it works (opens in a new tab)
Minimum reusable input
Implicit: 1,024 tokens. Explicit prefix: 256 tokens. Session cache has no minimum stated.Source for Minimum reusable input (opens in a new tab)
Lifetime & refresh
Implicit retention is dynamic. Responses explicit caching expires at expire_at, at most 7 days after creation; reuse does not extend expiration.Source for Lifetime & refresh (opens in a new tab)
How to enable it
Implicit caching is automatic. Responses explicit caching requires store=true and caching.type=enabled; reuse the stored response/prefix identifier.Source for How to enable it (opens in a new tab)
Usage fields
Inspect cached_tokens in the response usage details to distinguish hits from uncached input.Source for Usage fields (opens in a new tab)
Billing rules
Cached-input rates depend on prompt length. Implicit storage is free. Explicit storage is billed separately, rounded to natural hours.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Explicit caching replaces implicit caching for that request. The older Context API uses a different, sliding TTL; its default is not copied to Responses caching.Source for Limits & data boundaries (opens in a new tab)
Cache reads, writes & storage
ChargeApplies toRate / unit
Cache read1Text tokens · Prompt length [0, 256] K tokens$0.10 / 1M tokensPricing evidence (opens in a new tab)
Explicit cache storage2Caching$0.0083 / 1M tokens/hourPricing evidence (opens in a new tab)

Rate conditions

  1. standard non-audio rate table. Full prompt length selects the rate for the entire request; thinking tokens count toward output. Provider lists the Turbo family price, linked to the sole Turbo revision in the reviewed catalog. Finite pricing bounds do not override modality-specific API input limits.
  2. Prompt length [0, 256] K tokens. Explicit cache only; implicit storage is free. Partial natural hours billed as whole hours. Pricing-table units used; the cache guide uses a different generic unit description.

Observed 12 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 12 Oct 2026

BytePlus ModelArk · flex

dola-seed-2-1-turbo-260628
Documented

A cached-input rate is published for this service tier; this does not imply the standard endpoint’s explicit-storage controls.

Support evidence (opens in a new tab)
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Use the cached-input rate for this service tier and prompt-length band.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Not collected
Cache reads, writes & storage
ChargeApplies toRate / unit
Cache read1Text tokens · Prompt length [0, 256] K tokens$0.10 / 1M tokensPricing evidence (opens in a new tab)

Rate conditions

  1. flex non-audio rate table. Full prompt length selects the rate for the entire request; thinking tokens count toward output. Provider lists the Turbo family price, linked to the sole Turbo revision in the reviewed catalog. Finite pricing bounds do not override modality-specific API input limits.

Observed 12 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 12 Oct 2026

BytePlus ModelArk · batch

dola-seed-2-1-turbo-260628
Documented

A cached-input rate is published for this service tier; this does not imply the standard endpoint’s explicit-storage controls.

Support evidence (opens in a new tab)
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Use the cached-input rate for this service tier and prompt-length band.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Not collected
Cache reads, writes & storage
ChargeApplies toRate / unit
Cache read1Text tokens · Prompt length [0, 256] K tokens$0.10 / 1M tokensPricing evidence (opens in a new tab)

Rate conditions

  1. batch non-audio rate table. Full prompt length selects the rate for the entire request; thinking tokens count toward output. Provider lists the Turbo family price, linked to the sole Turbo revision in the reviewed catalog. Finite pricing bounds do not override modality-specific API input limits.

Observed 12 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 12 Oct 2026

Volcengine Ark (China) · Standard online; other family tariff modes are listed separately and require eligibility.

doubao-seed-2-1-turbo-260628
Documented
Support evidence (opens in a new tab)
How it works
Implicit prefix reuse; explicit Responses prefix/session caching.Source for How it works (opens in a new tab)
Minimum reusable input
Implicit: 2,048 tokens. Explicit prefix: 256 tokens; session: no minimum.Source for Minimum reusable input (opens in a new tab)
Lifetime & refresh
Implicit retention is dynamic. Explicit expire_at is at most seven days ahead and reuse does not renew it.Source for Lifetime & refresh (opens in a new tab)
How to enable it
Implicit is automatic. Explicit uses store=true, caching.type=enabled and previous_response_id.Source for How to enable it (opens in a new tab)
Usage fields
Reuse matching prefixes or stored conversation history. Implicit hits are not guaranteed.Source for Usage fields (opens in a new tab)
Billing rules
Implicit storage is free; explicit storage costs CNY 0.017 per 1M tokens/hour, rounded to natural hours. Hit-input and output tokens are separate charges.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Explicit cache history allows json_object, not json_schema; prefix creation cannot stream. Maintain thinking settings across reused history.Source for Limits & data boundaries (opens in a new tab)
Cache reads, writes & storage
ChargeApplies toRate / unit
Explicit cache storage · Input length [0, 256] (thousand tokens)1Volcengine Standard online family tariffCN¥0.017 / 1M tokens/hourPricing evidence (opens in a new tab)
Cached non-audio input · Input length [0, 256] (thousand tokens)1Volcengine Standard online family tariffCN¥0.60 / 1M tokensPricing evidence (opens in a new tab)
Cached non-audio input · Input length [0, 256] (thousand tokens)2Volcengine Low-latency online family tariffCN¥1.20 / 1M tokensPricing evidence (opens in a new tab)
Cached non-audio input · Input length [0, 256] (thousand tokens)3Volcengine Flex online family tariffCN¥0.60 / 1M tokensPricing evidence (opens in a new tab)
Cached non-audio input · Input length [0, 256] (thousand tokens)4Volcengine Batch inference family tariffCN¥0.60 / 1M tokensPricing evidence (opens in a new tab)

Rate conditions

  1. doubao-seed-2.1-turbo, Standard online; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
  2. doubao-seed-2.1-turbo, Low-latency online; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
  3. doubao-seed-2.1-turbo, Flex online; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.
  4. doubao-seed-2.1-turbo, Batch inference; request input length [0, 256] thousand tokens determines the whole-request rate. Family schedule does not establish dated-endpoint eligibility for every mode. Explicit storage is extra; Flex allows implicit cache only. Native CNY, no currency conversion.

Observed 15 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 15 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Inferred from the named native hosted API offering and cross-check of ByteDance and ByteDance-Seed public repositories. No matching downloadable checkpoint was found. This is not an explicit published weight-license declaration.

Classification evidence (opens in a new tab)

Sources reviewed