API pricing · Reporting
Rumble-2-Flash: check the paid plan behind €0.01/hour
The new open weights and hosted API have different controls. Check the plain-transcript restriction and monthly credit fee before turning an hourly rate into a budget.
Open weights are the new release, not a new API launch
Reson8's November 18 announcement opens the weights of Rumble-2-Flash, a 0.5B-parameter speech-recognition model covering 40+ languages. It does not mark the first availability of its hosted API: Reson8 announced that on November 3. For a creator choosing how to transcribe recordings, separate what you can run locally from what a managed endpoint actually returns.
The model card documents Transformers and vLLM use, including controls for digits, fillers and repeated words. That is a local-model interface, not a guarantee of matching hosted parameters. Running downloaded weights also leaves you responsible for runtime and compute; the open-weight announcement is not a quote for a managed API bill.
Plain text is not a timestamped transcript
On Reson8's hosted API, select model=rumble-2-flash explicitly. The current model documentation says omitting model selects resonant-1, not Flash. Flash supports Prerecorded, not Realtime or Turns, and returns a plain transcript. Requests with diarize, include_timestamps, include_words, phrases, custom_model_id, patterns, filler_mode or number_format are rejected. Do not infer those options from the local model card or features listed for Rumble-2.
If your meeting workflow needs speaker labels or word timing for editing, the cheaper Flash row does not describe an equivalent service. Check required output fields first, then compare compatible options. For a plain voice note, keep a check set with names, numbers and typical background noise. Record correction work as well as whether the endpoint accepted the file.
Price the paid month, not just the audio hour
The pricing page lists Flash Prerecorded at €0.0114 per audio hour on Launch and €0.01 on Scale. Launch costs €20 per month and Scale €100; the fee becomes credits usable across models and endpoints. The provider bills processed-audio seconds and warns that displayed hourly and per-minute rates are rounded. Listed prices exclude VAT.
Consider the hypothetical table: one new paid month containing only 1,000 audio hours of Flash transcription, without free or promotional credits. At the displayed rates, the usage value is about €11.40 on Launch and €10 on Scale. Each fits within its selected plan's credit-bearing fee, so the new-month outlay remains €20 or €100, respectively, rather than just the smaller usage value. These are modeled amounts, not bills from a test.
Build has monthly free credits and unused credits do not roll over, so this is not a ranking of every purchasing path. An existing customer's incremental cost can also differ because of their remaining credits and other usage. Confirm credit eligibility, needed concurrency and the completion deadline before buying. Reson8 positions Flash for testing and non-critical workloads; a lower hourly rate does not remove the task's output requirements.
1,000 hours: usage value versus a new paid month
Scroll sideways to read all columns.
| Paid plan | Monthly fee | Rate / audio h | Approx usage value | New month outlay |
|---|---|---|---|---|
| Launch | €20 | €0.0114 | €11.40 | €20 |
| Scale | €100 | €0.01 | €10.00 | €100 |
Sources & verification
Provider documentation supports the facts referenced below. Our explanations and recommendations are editorial interpretation.
- Reson8 on Hugging Face: Rumble-2-Flash open-weight announcement ↗Checked Nov 19, 2026
- Reson8/Rumble-2-Flash model card ↗Checked Nov 19, 2026
- Reson8 hosted speech-to-text model documentation ↗Checked Nov 19, 2026
- Reson8 pricing and billing FAQ ↗Checked Nov 19, 2026
- Reson8: Meet Rumble-2-Flash, November 3 API launch ↗Checked Nov 19, 2026
Corrections: hello@aiapicost.com