InvokeAI
InvokeAI 6.14 ships video generation, multi-GPU support, and six new model families
A side-by-side editorial comparison of Exa and Ollama — release velocity, themes, recent moves, and the top alternatives to consider.
Exa is pushing past search into autonomous web-research agents.
Exa has moved beyond its search-and-retrieval API into agentic territory. The headline change is Exa Agent — a research agent built on Exa's index and reachable via API — now joined by MCP availability for Agent and Connect. The underlying search product keeps maturing in parallel: auto-routing, people and company search, markdown-native content, and instant results.
Ollama's v0.34.x RC chain fixes a 90 GB speculative-decode memory explosion and opens thinking levels to the API.
Ollama is mid-cycle in a rapid v0.34.x release-candidate chain, with the bulk of work targeting MLX performance on Apple Silicon. The most significant recent fix resolved a speculative-decode memory regression that was pushing runner footprints past 90 GB and crashing kernels during long 98k-token contexts. Alongside that, the API surface expanded to expose thinking levels and model defaults — a direct response to the proliferation of reasoning-capable models.
Exa has moved beyond its search-and-retrieval API into agentic territory. The headline change is Exa Agent — a research agent built on Exa's index and reachable via API — now joined by MCP availability for Agent and Connect. The underlying search product keeps maturing in parallel: auto-routing, people and company search, markdown-native content, and instant results.
The arc runs from primitives to products: a fast index, then specialized verticals (people, companies), now an agent that composes them into end-to-end research. Bringing Agent and Connect to MCP signals Exa wants to be a retrieval backend inside other agent stacks, not just a standalone API.
Expect Exa to deepen the agent layer — structured research outputs and monitoring already appear in the changelog — and to lean on MCP distribution to embed inside third-party agents rather than compete for end users directly.
Ollama is mid-cycle in a rapid v0.34.x release-candidate chain, with the bulk of work targeting MLX performance on Apple Silicon. The most significant recent fix resolved a speculative-decode memory regression that was pushing runner footprints past 90 GB and crashing kernels during long 98k-token contexts. Alongside that, the API surface expanded to expose thinking levels and model defaults — a direct response to the proliferation of reasoning-capable models.
The consistent thread across this window is MLX investment: Ollama is iterating on Apple Silicon performance (Qwen 3.8 prompt speedups, KV buffer management, speculative decode stability) while simultaneously expanding its API to surface reasoning-model controls. The shared CLI/desktop first-run onboarding signals a deliberate push toward a broader, less technical user base. Ollama is building depth on Apple hardware while widening the top of the funnel.
A stable v0.34.x release is the immediate next step once the RC chain clears. After that, the thinking-level API field sets up first-party and third-party integrations to begin differentiating on reasoning-model configuration — watch for client libraries to start using it.
Other ai-assistants products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either Exa or Ollama.
InvokeAI 6.14 ships video generation, multi-GPU support, and six new model families
Copilot wires persistent memory into agentic security as it broadens its model roster and enterprise defaults.
Claude opens a developer plugin portal — platform play, not just a model.
Baseten moves beyond model hosting with built-in web search and Grounded Inference.
KServe v0.21.0 ships as the GA release of a cycle that turned the platform into a production LLM inference layer.
Poe's App Creator matures into a Claude-native platform for building and monetizing AI applications.
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
They serve adjacent needs but don't currently overlap on shipped themes. Ollama is currently shipping more aggressively (velocity 7.5 vs 6.3), with 1 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. Ollama is currently shipping more aggressively (velocity 7.5 vs 6.3), with 1 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other ai-assistants products to evaluate alongside.
Top Exa alternatives in ai-assistants are ranked by recent ship velocity. Browse the "Exa alternatives" section above for the current picks, or visit /alternatives/exa for the full list with editorial commentary on each.
Top Ollama alternatives in ai-assistants are ranked by recent ship velocity. Browse the "Ollama alternatives" section above for the current picks, or visit /alternatives/ollama for the full list with editorial commentary on each.