RunInfra · First-party API · LLM

ornith-1-5-35b

ornith-1-5-35b
Release announcedReleased on Aug 20, 2026

RunInfra released ornith-1-5-35b on Aug 20, 2026. Source: runinfra.ai · Last checked Nov 14, 2026

Notice

Release announced

Released on Aug 20, 2026

RunInfra announced this model on 2026-08-20.

Hosting route
First-party API
Affected scope
Not stated in the notice
Announced
Aug 20, 2026
First seen by ModelClock
Nov 11, 2026

What the provider published

runinfra.ai ↗

dings%2C-and-rerank-are-retired,children:Speech to text, embeddings, and rerank are retired}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:Parakeet TDT 0.6B v3, Qwen3 Embedding 0.6B, Qwen3 Embedding 8B, and Qwen3 Reranker 8B no longer serve.}), A call naming any of them returns ,_jsx(_components.code,{children:404 model_not_found}),. No hosted model answers ,_jsx(_components.code,{children:POST /v1/audio/transcriptions}),, ,_jsx(_components.code,{children:POST /v1/embeddings}),, or ,_jsx(_components.code,{children:POST /v1/rerank}), today. ,_jsx(_components.code,{children:GET /v1/models}), is the authoritative live list.]}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[A link to a removed model lands on the ,_jsx(_components.a,{href:https://runinfra.ai/inference-api`,children:`Model Library})]}), rather than on a dead page.]}),_jsx(Heading,{level:2,id:the-legacy-tool-dialect-is-refused,children:The legacy tool dialect is refused}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[_jsx(_components.code,{children:functions}), and ,_jsx(_components.code,{children:function_call}), return ,_jsx(_components.code,{children:400}), on DeepSeek V4 Flash.]}), The deprecated fields were previously accepted and silently discarded, so the definitions never reached the model and the reply came back as prose no tool parser could read. ,_jsx(_components.code,{children:tools}), and ,_jsx(_components.code,{children:tool_choice}), are the supported dialect. The refusal is declared per model, and DeepSeek V4 Flash is the model it was measured on.]}),_jsx(Heading,{level:2,id:nemotron-3-5-lightning-earns-its-cached-rate,children:Nemotron 3.5 Lightning earns its cached rate}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[_jsx(_components.code,{children:nemotron-3-5-lightning-30b}), bills cached input at $0.01 per million tokens.]}), A cached rate is published only where prefix-cache isolation has been verified on the served model. Nemotron’s attestation landed, so the rate its row carried is now the rate you are charged.]}),_jsx(Heading,{level:2,id:team-accounts,children:Team accounts}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:You can invite members, switch accounts, and leave an account.}), Members is a settings destination with two roles, owner and member. Billing and the commercial relationship stay with the owner.]})]}),\n,_jsx(Update,{label:August 23, 2026,description:a session header that keeps a conversation on the same warm prefix cache,tags:[API,Model APIs],id:august-23%2C-2026,children:_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[_jsx(_components.code,{children:x-session-id}), and ,_jsx(_components.code,{children:x-parent-session-id}), are first-class session identities.]}), Precedence is ,_jsx(_components.code,{children:x-session-id}),, then ,_jsx(_components.code,{children:x-parent-session-id}),, then ,_jsx(_components.code,{children:x-session-affinity}),, so a subagent stream carrying only its parent’s id shares the parent’s warm prefix cache. A request that sends none of the three is byte identical to before.]})}),\n,_jsxs(Update,{label:August 22, 2026,description:longer response ceilings, a larger share of a model's concurrency, a free tier that is not one request per second, and large idempotent replays,tags:[API,Model APIs,Reliability],id:august-22%2C-2026,children:[_jsxs(_components.p,{children:[_jsx(_components.strong,{children:Response time ceilings went up.}), Maximum request duration moved from 300 to 800 seconds, and the platform response ceiling from 240 to 740 seconds, after a customer running long non-streaming completions failed 9 of 10 calls at the old ceiling. The time allowed for a streaming response to begin moved from 60 to 180 seconds; a non-streaming response has the full ceiling.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:A paying workspace gets half of a model’s concurrency slots, not a quarter.}), Never the whole model, so a second paying tenant can still get in.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:The free tier’s requests per minute moved from 60 to 10,000.}), One request per second is below what a single agentic client idles at. The published tier table now matches what is enforced.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:A large response can be idempotently replayed.}), The replay record ceiling went from 512 KiB to 6 MiB. Above the old ceiling, a retry with the same ,_jsx(_components.code,{children:Idempotency-Key}), used to get a permanent ,_jsx(_components.code,{children:422}),.]})]}),\n,_jsxs(Update,{label:August 21, 2026,description:spend raises the tokens-per-minute budget, and a token meter that had been counting bytes,tags:[Billing,API,Model APIs],id:august-21%2C-2026,children:[_jsxs(_components.p,{children:[_jsx(_components.strong,{children:Spend tiers raise the tokens-per-minute budget.}), Starter keeps the model default, unchanged for every existing workspace, and Scale and Pro raise it. Qualification is by cumulative amount paid. Concurrency is deliberately not tiered: the token meter is a fairness budget, and the concurrency share is the capacity guard. The dashboard draws all three tiers with your current one marked.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:The token meter stopped counting bytes.}), It was measuring bytes rather than tokens, so every workspace’s per-minute budget was effectively a quarter of the published figure.]})]}),\n,_jsxs(Update,{label:August 20, 2026,description:Ornith 1.5 35B is public, and every refusal says when to retry,tags:[Model APIs,Models,API],id:august-20%2C-2026,children:[_jsx(Heading,{level:2,id:ornith-1-5-35b-is-public,children:Ornith 1.5 35B is public}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[_jsx(_components.code,{children:ornith-1-5-35b}), joins the hosted models.]}), A 262,144 token context window and a 32,768 token output budget, priced at $0.10 per million input tokens, $0.01 per million cached input tokens, and $0.40 per million output tokens. It accepts text and image input, and it caches prefixes.]}),_jsx(Heading,{level:2,id:every-refusal-says-when-to-retry,children:Every refusal says when to retry}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[A paused model and a rate limit both carry ,_jsx(_components.code,{children:Retry-After}),.]}), A paused refusal sets it to the distance to the next availability check. A ,_jsx(_components.code,{children:429}), sets it from the state of the limiter that refused you, in seconds and in milliseconds. A capability or parameter refusal carries no retry header, because the same request will not succeed later.]})]}),\n,_jsxs(Update,{label:August 18, 2026,description:DeepSeek V4 Pro is public, image input on Qwen3.8 27B, one hosted checkout, and a measured cache hit rate on model pages,tags:[Model APIs,Models,Billing],id:august-18%2C-2026,children:[_jsx(Heading,{level:2,id:deepseek-v4-pro-is-public,children:DeepSeek V4 Pro is public}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:DeepSeek V4 Pro joins the hosted models.}), A 1,048,576 token context window and a 32,768 token output budget, priced at $0.60 per million input tokens, $0.03 per million cached input tokens, and $1.90 per million output tokens. Text input, and it caches prefixes.]}),_jsx(Heading,{level:2,id:image-input-on-qwen3-8-27b,children:Image input on Qwen3.8 27B}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[_jsx(_components.code,{children:qwen3-8-27b}), launched with images as inline data URLs.]}), Accepted input is now a per-model fact, and it drives each model page’s Accepted input row, the refusal a text-only model returns, and the input modalities we publish to catalogs. This image guidance was superseded on October 25, 2026. Read the ,_jsx(_components.a,{href:/api-reference/chat-completions#three-things-a-request-can-be-refused-for,children:current image URL support}),.]}),_jsx(Heading,{level:2,id:one-checkout,children:One checkout}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:Every purchase goes through hosted checkout.}), The purchase carries a durable identifier for that attempt, returns to the page you started from, verifies the grant on return, and survives a back button. Billing details are collected at checkout.]}),_jsx(Heading,{level:2,id:model-pages-publish-a-measured-cache-hit-rate,children:Model pages publish a measured cache hit rate}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:A model with a cached input price publishes the share of input tokens actually served from cache.}), It is derived from settled billing over a trailing day, and it is absent rather than estimated when the measurement cannot be made.]})]}),\n,_jsxs(Update,{label:August 16, 2026,description:model discovery publishes context and prices, and a retention statement that matches what is stored,tags:[API,Privacy,Model APIs],id:august-16%2C-2026,children:[_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[_jsx(_components.code,{children:GET /v1/models}), and ,_jsx(_components.code,{children:GET /v1/models/{model}}), publish ,_jsx(_components.code,{children:context_length}), and a ,_jsx(_components.code,{children:pricing}), object.]}), Rates are in US dollars per token, and a cached rate appears only for a model whose prefix-cache isolation is attested. A newly registered model publishes both with no code change.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:The data retention statement is exact.}), The policy said content is never stored, while the idempotent replay cache holds non-streaming response bodies for 24 hours by design. It now states that window and what it is bound to, notes that streaming is exempt and that a keyless request leaves no stored content, and every model page carries a Data retention row from the same source these docs read. See ,_jsx(_components.a,{href:/security/data-retention,children:Data retention}),.]})]}),\n,_jsxs(Update,{label:August 15, 2026,description:Idempotency-Key covers streaming, cached input pricing, credit top-ups that cannot shrink or double charge, and a named milestone when you cross one,tags:[API,Billing,Credits,Model APIs],id:august-15%2C-2026,children:[_jsx(Heading,{level:2,id:idempotency-key-now-covers-streaming,children:Idempotency-Key now covers streaming}),_jsxs(_components.p,{children:[_jsxs(_components.strong,{children:[A dropped stream retried with the same ,_jsx(_components.code,{children:Idempotency-Key}), no longer generates or charges twice.]}), Streaming chat completions used to ignore the header. They now record the key before the model runs, so a duplicate arriving while the stream is open gets ,_jsx(_components.code,{children:409 idempotency_conflict}), and starts no second generation, and a duplicate arriving after the original settled gets the terminal usage and cost as JSON with ,_jsx(_components.code,{children:X-RunInfra-Idempotent-Replay: true}),. We do not store the tokens we deliver, so the text itself is never replayed. A streaming key is held for the response deadline plus a short settlement grace, 270 seconds at the outside, then clears on its own. The bound follows the response ceiling, so since the August 22, 2026 change it is 770 seconds.]}),_jsxs(_components.p,{children:[This changes behavior for a client that sends one constant key: give each logical request its own key, and reuse a key only to retry that same request. The contract is in ,_jsx(_components.a,{href:/api-reference/idempotent-retries,children:Idempotent retries}),.]}),_jsx(Heading,{level:2,id:cached-input-is-priced-separately,children:Cached input is priced separately}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:A repeated prompt prefix bills at a cached input rate, on the models where cache isolation is proven.}), On an attested model, cache isolation is per workspace: a prefix cached for one workspace is never served to another. A model earns a published cached rate only once that isolation has been verified on the served model, with an identical prefix under a different workspace’s cache identity confirmed to return no cached tokens. DeepSeek V4 Flash, ,_jsx(_components.code,{children:qwen3-8-27b}),, and Qwen3.8 2.4T A95B are attested today. A model with no attestation publishes no cached price and bills cached input at its standard input rate.]}),_jsx(Heading,{level:2,id:credit-top-ups,children:Credit top-ups}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:Promotion codes no longer apply to credit top-ups.}), Credits are stored value, so a discount code reduced the credits you received by exactly what it took off the price: it appeared to work and bought you nothing. The amount you select is now the amount you are charged and the amount that lands on your balance.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:A reload cannot charge you twice.}), A top-up carries a durable identifier for that attempt, so a refresh, a back button, a second submit, or a return from a bank verification screen resumes the same payment instead of starting another.]}),_jsxs(_components.p,{children:[_jsx(_components.strong,{children:A confirmation that does not settle is reported as unknown, never as declined.}), When confirmation does not finish inside its bound, the screen says the result is not y

Read from the provider's text; the quote is the provider's exact lines.

Revision history

Loading…
Loading revision history…
Raw history JSON

Add a badge to your README

ornith-1-5-35b shutdown date

Shows this model's current status and follows new notices from the provider. Unknown models show “not tracked”, so a README never breaks.

Markdown
[![ornith-1-5-35b shutdown date](https://model-clock.com/badge/runinfra/ornith-1-5-35b.svg)](https://model-clock.com/models/runinfra/ornith-1-5-35b)
HTML
<a href="https://model-clock.com/models/runinfra/ornith-1-5-35b"><img src="https://model-clock.com/badge/runinfra/ornith-1-5-35b.svg" alt="ornith-1-5-35b shutdown date"></a>

Prefer Shields.io styles? Use the endpoint badge with https://model-clock.com/badge/runinfra/ornith-1-5-35b.json.