Multi-Provider Parallel Generation

Multi-Provider Parallel lets a single bulk run fan out across several AI providers at the same time — for example Gemini, Claude, and DeepSeek all generating concurrently. Because every provider has its own separate rate-limit pool, running them together clears far more work per minute than hammering one provider's limit.

It's the fastest option for the live generation path (results appear in real time). For the cheapest way to process huge batches, see Batch Generation below — the two are complementary.

How it works: weighted lanes

You choose which providers join the run and give each a weight — the number of concurrent "lanes" it gets. The app expands those weights into a pool of lanes, where each lane is a fixed (provider, model, key). Every lane pulls the next row from the queue, generates it, and immediately pulls the next — so faster providers naturally do more of the work.

With one Gemini key you get one lane. Add weights across providers and you get many — a Gemini ×3 + Claude ×1 + DeepSeek ×2 pool runs 6 articles concurrently instead of one at a time.

Setting it up

Open Settings → API Keys and turn on Multi-Provider Parallel. Tick the providers you want in the pool (only providers with a saved key appear), then set each one's lane weight with the − / + steppers. The card shows the live total.

Then just run a generation as usual — when the pool is on, the work is automatically distributed across your lanes.

Where it applies

For metadata and outlines the output is uniform (JSON / heading structure), so mixing providers has no downside — it's pure speed. For full article bodies, expect some variation in voice and style between articles, since each is written by a different model — so you'll get an explicit confirmation before a mixed-provider drafting run.

Notes & behavior

• Bring your own keys. A provider only appears in the pool if it has a saved key (single key or an enabled multi-key entry). Each provider uses the model you've selected for it in Settings.
• Resilient. If one provider hits its rate limit, its rows are marked and the other lanes keep going — re-run to retry, and those rows can route to a provider that still has headroom.
• Off by default. When disabled, generation uses your single selected provider exactly as before.
• Per-user setting. Your pool configuration is saved to your browser and reused across runs.

Batch vs Multi-Provider Parallel

These are the two ways to scale a big run — pick based on whether you need results now or cheapest:

They're mutually exclusive per run — choose one. Batch is offered as a toggle in the Generate-SEO and Outline actions; Multi-Provider Parallel applies whenever the pool is enabled and you run live.