404 model_not_found: qwen/qwen3.6-27b (Groq)
This is Groq's most recent chat-model retirement, and the one with the least warning on the deprecations page: 404 The model `qwen/qwen3.6-27b` does not exist or you do not have access to it has been the answer for free and developer-tier accounts since 14 September 2026. The model had only been on Groq's list since the summer, when Groq named it as a replacement for llama-3.3-70b-versatile and llama-4-scout, so a fair number of apps that followed those notices were sent to an id that lasted two months.
Last verified November 18, 2026 against Groq deprecations, Groq model list · 5 min read
| Provider | Groq |
| HTTP status | 404 Not Found |
| Error code | model_not_found |
| Message | The model `qwen/qwen3.6-27b` does not exist or you do not have access to it. |
| Announced / shut down | undated notice on Groq's page / 14 September 2026 |
| When it happens | Since 2026-09-14, on any free or developer-tier request that names this model |
| Can you wait it out? | No — the model was retired; the request fails until you change the model or the provider |
- Fastest: change the model string to
qwen/qwen3.8-27bon Groq. Different model, so re-check your prompts. - Least code change: keep
qwen/qwen3.6-27band change only the base URL; Inference APIs accepts the id and answers with Qwen3.8 Flash. - So it does not happen again: read the model id from configuration and add a fallback for 404 and 429.
What happened
Groq's deprecations page lists the retirement under 14 September 2026 with qwen/qwen3.8-27b as the replacement, described as the direct successor: a 27B multimodal model with the same 131K window, thinking and instruct modes, tunable reasoning effort, tool use and JSON mode. The notice applies to free and developer-tier usage; enterprise contracts are not affected. Unlike Groq's other 2026 retirements, the page gives no announcement date for this one. A week later, on 21 September 2026, Groq also retired its own groq/compound and groq/compound-mini systems with no replacement listed.
First, find where the id is pinned
The id is often in more places than the one that errored: a default in a settings file, an environment variable on the server, a second call for summarisation or titles, a framework default.
grep -rn -E "qwen/qwen3.6-27b|qwen3.6-27b" . --include='*.py' --include='*.ts' --include='*.js' --include='*.json' --include='*.yaml' --include='*.yml' --include='*.toml' --include='*.env*'
# and the places grep cannot see: environment variables on the server, a hosted secrets store, the provider dashboard's default modelOption 1 — switch to Groq's recommended replacement
This is the easy case: qwen/qwen3.8-27b is the same family one release on, with the same window and the same controls, so most prompts carry over. Still worth a test pass:
- Thinking defaults may differ between the two releases; if you relied on instruct mode, set it explicitly rather than assuming the default.
- Output wording changes with every Qwen release; few-shot parsers and strict output formats need a re-run.
- It is the second hop in a year for apps that arrived here from the Llama 3.3 70B or Scout retirements, so put the id in configuration this time.
resp = client.chat.completions.create(
model="qwen/qwen3.8-27b", # was "qwen/qwen3.6-27b"
messages=[{"role": "user", "content": "Hello"}],
)Option 2 — keep the old id, change the base URL
Inference APIs accepts qwen/qwen3.6-27b and qwen3.6-27b as aliases and serves them with Qwen3.8 Flash, the current Qwen 3.8 generation: a 1,000,000-token window, thinking on by default with reasoning_effort: "none" to switch it off, tool calling and JSON output, at $0.15 / $0.50 per million tokens with cached input at $0.02.
from openai import OpenAI
client = OpenAI(
base_url="https://api.inferenceapis.com/v1", # was https://api.groq.com/openai/v1
api_key=os.environ["INFERENCE_API_KEY"],
)
resp = client.chat.completions.create(
model="qwen/qwen3.6-27b", # the retired Groq id, accepted as an alias
messages=[{"role": "user", "content": "Hello"}],
)Qwen3.8 Flash is text-only here, so if you were sending images to the 27B on Groq, use Qwen/Qwen3-VL-235B-A22B-Instruct for those requests instead. Prices and capabilities: Qwen3.8 Flash here.
Option 3 — stop pinning one provider and one id
Put the model id in configuration and let a 404 or 429 fall through to a second provider. A retirement then degrades to a log line instead of an outage, and you choose when to migrate. LiteLLM's router does this in a few lines; the same shape exists in most frameworks.
# LiteLLM proxy config: Groq first, Inference APIs when Groq answers 404 or 429
model_list:
- model_name: chat
litellm_params:
model: groq/openai/gpt-oss-120b
api_key: os.environ/GROQ_API_KEY
- model_name: chat-fallback
litellm_params:
model: inferenceapis/openai/gpt-oss-120b
api_key: os.environ/INFERENCEAPIS_API_KEY
router_settings:
fallbacks: [{ "chat": ["chat-fallback"] }]The deprecations tracker lists every Groq retirement with its date and replacement, so you can see the next one coming.
Frequently asked questions
Is qwen/qwen3.8-27b the same as Qwen3.8 Flash?
Both are Qwen 3.8 releases, but different sizes and builds: Groq's is the 27B dense model, Qwen3.8 Flash is Alibaba's fast tier with a 1M window. Prompts carry over; expect different wording and test before you stop looking.
Why was a two-month-old model retired?
Groq's page does not say. Its list has turned over a Qwen slot roughly every few months since 2025 (qwen-qwq-32b, qwen/qwen3-32b, qwen/qwen3.6-27b, now qwen/qwen3.8-27b), so treat any Qwen id on Groq as something to read from configuration.
How do I turn thinking off on Qwen3.8 Flash here?
Send reasoning_effort: "none". The response then has no reasoning_content field and the answer comes back in one piece.
Where Inference APIs fits
Option 2 above keeps the old id working with a base-URL change: Inference APIs accepts qwen/qwen3.6-27b and serves it with Qwen3.8 Flash, the same model generation Groq moved to, at $0.15 / $0.50 per million tokens with cached input at $0.02 and no daily cap.
Something changed or wrong? Tell us and we will re-verify the entry.
