---
title: "Performance"
description: "817 posts about Performance, summarised, each linking to the original."
---

# Performance
> 817 posts about Performance, summarised, each linking to the original.

## Articles

### [vLLM V0 to V1: Correctness Before Corrections in RL](https://yomu.fyi/post/vllm-v0-to-v1-correctness-before-corrections-in-rl.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Rafael Pardinas, Ehsan Kamalloo
- Published: May 6, 2026
- Type: Problem & solution

PipelineRL migrated its rollout inference engine from vLLM V0 (0.8.5) to V1 (0.18.1), but trainer-side logprobs, policy ratios, KL, entropy, clip rate, and reward diverged from the V0 reference. The investigation separated semantic mismatch, inference-path mismatch, and objective mismatch, then prioritized backend parity before changing the RL objective. Fixes included processed rollout logprobs, explicit V1 runtime settings disabling prefix caching and async scheduling, a V0-like inflight weight-update path, and an fp32 lm\_head for the final projection. With those changes, the final V1 run tracked the V0 trajectory across the reported metrics, while processed\_logprobs alone did not resolve the training gap.


### [Improved debugging for Expo apps with the React Native SDK](https://yomu.fyi/post/improved-debugging-for-expo-apps-with-the-react-native-sdk.md)
- Company: [Sentry](https://yomu.fyi/company/sentry.md)
- Author: Aleksandr Pantiukhov
- Published: May 6, 2026
- Type: Announcement

Expo applications generate approximately 75% of the total event volume received from React Native applications in Sentry. In response, Sentry updated its React Native SDK starting in version 8.10 to improve Expo debugging and performance monitoring workflows. The update automatically enriches events with Expo constants and over-the-air update context, including channels, runtime versions, and emergency launch warnings when an OTA update fails. Developers can track remote EAS Build lifecycle failures and successes using dedicated hooks configured in package.json. Performance instrumentation now supports span generation for Expo Router route prefetching as well as asset and image loading through wrapper functions around expo-image and expo-asset.


### [Adding Benchmaxxer Repellant to the Open ASR Leaderboard](https://yomu.fyi/post/adding-benchmaxxer-repellant-to-the-open-asr-leaderboard.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Eric Bezzam, Steven Zheng, Eustache Le Bihan, Sergio Bruccoleri, Jeanine Sinanan-Singh, Casey Ford, Guanbo Wang, Yukai Huang, Ke Li, Yufeng Hao, Liao Xiaoling
- Published: May 6, 2026
- Type: Announcement

The Open ASR Leaderboard is adding private English speech-recognition datasets from Appen Inc. and DataoceanAI to evaluate scripted and conversational speech across several accents. To reduce benchmark-specific optimization (“benchmaxxing”) and test-set contamination, the datasets remain private, while the default Average WER continues to use public datasets only. A toggle exposes the private data, with macroaverages for overall, scripted, conversational, American, and non-American subsets; individual split scores are withheld to discourage provider- or accent-specific optimization. The leaderboard uses a shared Hub dataset, normalized transcripts and outputs, and open-source evaluation code, while private evaluation is run after public-set verification when models are submitted.


### [Agentic analytics starts with query-ready data: the write-side cost of Snowflake vs. ClickHouse](https://yomu.fyi/post/agentic-analytics-starts-with-query-ready-data-the-write-side-cost-of.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: Tom Schreiber, Lionel Palacin
- Published: May 6, 2026
- Type: Benchmark

Agentic analytics workflows introduce bursts of concurrent, low-latency analytical queries over continuously ingested data. Analytical query efficiency relies on columnar storage and chunk-level pruning, which require contiguous, sorted data to avoid full-table scans. Rather than evaluating warehouses solely at query runtime, this benchmark compares the write-side cost of maintaining query-ready data in ClickHouse and Snowflake during continuous ingestion at approximately one million rows per second. Snowflake relies on post-write clustering to organize data, whereas ClickHouse constructs sorted layouts on the write path and refines them via background merges. Across 100B, 200B, and 300B row checkpoints on the ClickBench dataset, ClickHouse achieved a query-ready layout at 22 times lower cost and delivered 28 times better write-side cost-performance.


### [Realtime or Pipelines? How to choose the right tool](https://yomu.fyi/post/realtime-or-pipelines-how-to-choose-the-right-tool.md)
- Company: [Supabase](https://yomu.fyi/company/supabase.md)
- Author: eduardo\_gurgel, riccardo\_busetti
- Published: May 5, 2026
- Type: Explainer

Supabase Realtime and Supabase Pipelines both read PostgreSQL changes through logical replication, but they target different jobs. Realtime delivers live updates and communication to browsers and apps through Broadcast, Presence, and Postgres Changes; its Postgres Changes path checks Row Level Security for subscribed users, uses temporary replication slots, and offers best-effort delivery without replay after disconnects. Pipelines is a managed change-data-capture system that performs an initial sync, then replicates selected publication data in batches to supported analytical destinations through a persistent replication slot, retaining progress and retrying transient failures. The choice depends on the destination and delivery needs: use Realtime for low-latency user experiences that can tolerate gaps, Pipelines for durable analytical copies with at-least-once processing and idempotent destination writes, and both when an application needs both capabilities.


### [Boosting multimodal inference performance by >10% with a single Python dictionary](https://yomu.fyi/post/boosting-multimodal-inference-performance-by-10-with-a-single-python-d.md)
- Company: [Modal](https://yomu.fyi/company/modal.md)
- Published: May 4, 2026
- Type: Problem & solution

Multimodal vision-language models pose new challenges for inference engines like SGLang. Profiling with py-spy revealed that the scheduler’s process\_input\_requests consumed about 13% of CPU time, with a quarter of that spent on repeated calls to torch.UntypedStorage.\_new\_shared\_cuda for CUDA IPC handle book-keeping. The solution was a simple Python dictionary cache for pool handles, eliminating redundant opens. On a Qwen2.5-VL-3B-Instruct benchmark with a single H100, throughput increased 16.2% from 22.2 to 25.7 requests per second, mean time-to-first-token dropped 13.2%, and mean time-per-output-token dropped 17.2%. The improvement is merged in SGLang v0.5.10 and applies to any multimodal model using SGLang’s CUDA IPC transport.


### [Gala supercharges analytics performance with ClickHouse on AWS](https://yomu.fyi/post/gala-supercharges-analytics-performance-with-clickhouse-on-aws.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: ClickHouse
- Published: May 4, 2026
- Type: Problem & solution

Blockchain gaming platform Gala struggled to ingest and query expanding volumes of telemetry and blockchain data on its Databricks-based data infrastructure, resulting in multi-minute query times and heavy engineering maintenance overhead. To solve these bottlenecks, the company migrated its data platform to ClickHouse on AWS, ingesting data through Kafka, Airbyte, Amazon S3, and Fivetran. Gala leveraged ClickHouse's indexing capabilities alongside saved queries and API endpoints to power company-wide Metabase business intelligence dashboards for non-technical teams. Following the migration completed in December 2024, data capacity expanded from 3 TB to 9 TB while query times dropped from minutes to sub-second on previously unoptimized tables. The switch also lowered initial infrastructure costs by 30 percent, with plans underway to implement ClickHouse ClickPipes for improved data processing pipelines.


### [Optimizing ML Workload Network Efficiency (Part I): Feature Trimmer](https://yomu.fyi/post/optimizing-ml-workload-network-efficiency-part-i-feature-trimmer.md)
- Company: [Pinterest](https://yomu.fyi/company/pinterest.md)
- Author: Pinterest Engineering
- Published: May 1, 2026
- Type: Problem & solution

Pinterest deployed a root-leaf architecture for online machine learning serving to decouple CPU-based feature preprocessing from GPU-based inference. Because the root cluster retrieved a union of features across models and fanned out the full set to every leaf, network bandwidth became a severe bottleneck. After initial tests with fbthrift LZ4 compression yielded a 20% bandwidth reduction at the cost of CPU and latency overhead, the team implemented Feature Trimmer. This mechanism uses exported model signatures as the source of truth to prune unused features on the root before transmission. By dispatching only required features per model, the system reduced network bandwidth, decreased latency, and enabled cluster rightsizing that saved over $4M annually.


### [Granite 4.1 LLMs: How They’re Built](https://yomu.fyi/post/granite-4-1-llms-how-they-re-built.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Yousaf Shah
- Published: Apr 29, 2026
- Type: Explainer

Granite 4.1 is a family of dense decoder-only language models in 3B, 8B, and 30B sizes, trained on approximately 15T tokens. Its five-phase pre-training pipeline moves from broad web-scale data to progressively curated mixtures, increases math and code emphasis, and extends context from 4K to 512K tokens through staged long-context training. Supervised fine-tuning uses about 4.1M curated samples evaluated with LLM-as-Judge and rule-based checks, while reinforcement learning uses on-policy GRPO with DAPO loss. The 8B instruct model matches or surpasses the earlier Granite 4.0-H-Small 32B-A9B MoE model despite its simpler dense architecture and fewer parameters, and the Granite 4.1 models are released under Apache 2.0.


### [Comparing ClickHouse versions with clickhousectl](https://yomu.fyi/post/comparing-clickhouse-versions-with-clickhousectl.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: Mark Needham
- Published: Apr 29, 2026
- Type: Tutorial

Evaluating query performance differences across ClickHouse versions previously required managing Docker containers or manually downloading historical binaries from GitHub releases. The clickhousectl command-line interface simplifies this process by downloading specific ClickHouse binaries, running multiple background server instances concurrently with automatic port assignment, and executing queries across local environments. Users can copy table schemas between local instances by piping raw output from SHOW CREATE TABLE queries and replicate datasets using ClickHouse's remote table function. In a benchmark comparing ClickHouse 25.12 and 26.3 on an S3-backed Parquet dataset, version 26.3 introduced a Parquet metadata cache that reduced repeat query times from roughly 9 seconds to 1–2 seconds. System tables like system.server\_settings and system.metrics enable operators to inspect cache configurations, eviction policies, and memory consumption directly.


### [Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents](https://yomu.fyi/post/introducing-nvidia-nemotron-3-nano-omni-long-context-multimodal-intell.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Tuomas Rintamaki, Amala Sanjay Deshmukh, Nabin Mulepati, Collin McCarthy, Pritam Biswas, Arushi Goel, Alexandre Milesi, Danial Mohseni Taheri, Kateryna Chumachenko, Isabel Hulseman, Zhehuai Chen, Karan, Tao, Leili Tavabi, Ehsan Hosseini Asl
- Published: Apr 28, 2026
- Type: Announcement

NVIDIA Nemotron 3 Nano Omni is introduced as an omni-modal understanding model for document analysis, multiple-image reasoning, automatic speech recognition, long audio-video understanding, agentic computer use, and general reasoning. It extends the Nemotron multimodal line from vision-language processing to text, image, video, and audio, targeting long, dense inputs such as 100-plus-page documents and mixed audio-visual evidence. Its unified encoder-projector-decoder combines a Nemotron 3 Nano 30B-A3B hybrid Mamba-Transformer MoE backbone with C-RADIOv4-H and Parakeet-TDT-0.6B-v2 encoders; dynamic visual resolution, Conv3D tubelets, and Efficient Video Sampling support multimodal context compression. The post reports leading or top results across document, video, audio-video, voice, and ASR benchmarks, alongside up to 9x higher throughput and 2.9x single-stream reasoning speed on multimodal use cases, and provides BF16, FP8, and NVFP4 checkpoints.


### [Formatting an entire 25 million line codebase overnight: the rubyfmt story](https://yomu.fyi/post/formatting-an-entire-25-million-line-codebase-overnight-the-rubyfmt-st.md)
- Company: [Stripe](https://yomu.fyi/company/stripe.md)
- Author: Fable Tales
- Published: Apr 28, 2026
- Type: Problem & solution

Stripe’s 25-million-line Ruby codebase needed a true autoformatter; before rubyfmt, no industry-wide Ruby formatter existed, and earlier tools had crashed on its files. rubyfmt began as a zero-configuration open-source project motivated by formatting disputes, Ruby’s slow startup, and the uncertainty faced by engineers arriving from other languages. To meet a 100ms target, it initially used a hand-authored C launcher and later moved formatting into Rust, with ripper parse trees first passing through JSON and eventually being deserialized directly from Ruby VALUE objects. Stripe adopted per-file opt-in, expanded coverage cautiously, and formatted the codebase on a Saturday in 2024; the excerpt reports that all 42-million-line Ruby codebase is now formatted. In 2025, migrating from ripper to Prism removed the need to link a Ruby VM and made rubyfmt smaller and dramatically faster.


### [ClickHouse at Google Cloud Next '26](https://yomu.fyi/post/clickhouse-at-google-cloud-next-26.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: ClickHouse
- Published: Apr 27, 2026
- Type: Announcement

ClickHouse announced several product updates and integrations designed to deepen native compatibility with Google Cloud. Migration of ClickHouse Cloud to Google Axion ARM-based processors demonstrated a 30 to 55 percent query speed increase and reduced data load times from 64 to 26 seconds on ClickBench benchmarks. An integration with Google Lakehouse Runtime Catalog enables querying Iceberg tables on Google Cloud Storage via the DataLakeCatalog engine without data movement. Furthermore, Bring Your Own Cloud reached general availability on Google Cloud Marketplace, and ClickPipes added an unordered mode using Pub/Sub notifications for GCS ingestion. Google Antigravity integrated ClickHouse through an MCP server to translate natural language into SQL and generate interactive React charts.


### [How to build scalable web apps with OpenAI's Privacy Filter](https://yomu.fyi/post/how-to-build-scalable-web-apps-with-openai-s-privacy-filter.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: yuvraj sharma, Freddy Boulton, Abubakar Abid
- Published: Apr 27, 2026
- Type: Tutorial

The article presents three web applications built around OpenAI’s open-source Privacy Filter, a PII detector that labels eight categories in one forward pass over a 128k-token context. Document Privacy Explorer extracts PDF or DOCX text and returns category-labeled spans, while Image Anonymizer combines Tesseract word boxes with character offsets to produce editable pixel redaction rectangles. SmartRedact Paste replaces detected spans with category placeholders, stores public and token-gated views, and evicts expired pastes in a daemon thread. Across the apps, gradio.Server’s @server.api routes model work through Gradio’s queue and gradio\_client, while plain FastAPI routes serve custom HTML, views, and lightweight lookups. The implementations keep browser-side filtering, canvas edits, and PNG export out of server round-trips, and the paste service is about 200 lines including storage.


### [DeepSeek-V4: a million-token context that agents can actually use](https://yomu.fyi/post/deepseek-v4-a-million-token-context-that-agents-can-actually-use.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: ben burtenshaw
- Published: Apr 24, 2026
- Type: Explainer

DeepSeek V4 is presented as an open model family aimed at long-running agentic workloads, with Pro and Flash MoE checkpoints offering 1M-token context windows. Its central architecture alternates Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA), compressing KV entries by 4x and 128x respectively, while using FP8/BF16 storage and FP4 indexing. The post-training design preserves reasoning across user turns when tool calls are present, introduces a dedicated XML tool-call schema, and relies on DSec, a Rust sandbox platform for RL rollouts. At 1M tokens, Pro uses 27% of V3.2's single-token FLOPs and 10% of its KV cache, while Flash uses 10% and 7%; V4-Pro-Max reports 80.6% on SWE Verified and 0.59 MRCR 8-needle retrieval at 1M.


### [What's New in pg\_clickhouse - JSONB Support, SQL value functions, Streaming, and more](https://yomu.fyi/post/what-s-new-in-pg-clickhouse-jsonb-support-sql-value-functions-streamin.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: David Wheeler
- Published: Apr 24, 2026
- Type: Announcement

Recent releases of pg\_clickhouse, a Postgres foreign data wrapper for ClickHouse, introduce query pushdown enhancements and streaming result handling. Version 0.1.10 adds pushdown for JSONB accessor operators and extraction functions by mapping them to ClickHouse sub-column syntax and toJSONString calls outside SELECT clauses. In version 0.2.0, date and time functions such as CURRENT\_DATE, CURRENT\_TIMESTAMP, and clock\_timestamp() push down to ClickHouse while honoring the active Postgres session time zone and precision settings. To address memory pressure and out-of-memory risks from large foreign queries, the extension adds query result streaming to the HTTP driver with a default batch buffer of approximately 50MB. Container benchmarking against the NYC taxi dataset showed streaming reduced peak memory consumption from over 600 MiB to under 86 MiB.


### [Do you still need Elasticsearch for log analytics? ClickHouse says no.](https://yomu.fyi/post/do-you-still-need-elasticsearch-for-log-analytics-clickhouse-says-no.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: Tom Schreiber, Lionel Palacin
- Published: Apr 23, 2026
- Type: Benchmark

ClickHouse introduced a redesigned full-text search powered by inverted indexes that feeds matching documents directly into its vectorized analytical engine for observability and log analytics workloads. To evaluate this capability, the open-source versions of ClickHouse and Elasticsearch were benchmarked on single-node AWS EC2 instances across synthetic OpenTelemetry log datasets scaling up to 50 billion rows. The benchmark evaluated nine representative query patterns combining text search with filtering, grouping, counting, and time-series aggregations under both cold and hot cache conditions. ClickHouse achieved a roughly fivefold smaller storage footprint on disk while running cold queries four to six times faster and hot queries approximately 1.7 to 2.6 times faster than Elasticsearch. Furthermore, ClickHouse ingested the 50-billion-row dataset in under four hours out of the box, whereas the tuned Elasticsearch setup required approximately five days.


### [Are LLM agents good at join order optimization?](https://yomu.fyi/post/are-llm-agents-good-at-join-order-optimization.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Eric Liang, Ryan Marcus, Sid Taneja, Yuhao Zhang
- Published: Apr 22, 2026
- Type: Benchmark

Join ordering is difficult because relational optimizers choose among exponentially many plans and depend on potentially inaccurate cardinality estimates. Databricks and UPenn researchers built an LLM agent that uses structured outputs to generate valid join reorderings and executes candidates through a tool returning runtime and intermediate subplan sizes. It can test up to 50 iterations, while the evaluation tested 15 rollouts per query across 113 Join Order Benchmark queries after duplicating each dataset row ten times. With a frontier model, the best tested orders improved geometric-mean latency by 1.288 times and reduced P90 latency by 41% versus the standard Databricks optimizer. A five-way example links its gain to LIKE predicates that are exceptionally difficult for cardinality estimators.


### [Frontend Engineering at Palantir: Engineering Multilingual Collaboration](https://yomu.fyi/post/frontend-engineering-at-palantir-engineering-multilingual-collaboratio.md)
- Company: [Palantir](https://yomu.fyi/company/palantir.md)
- Author: Palantir
- Published: Apr 22, 2026
- Type: Explainer

Palantir describes the frontend work behind adding read-only translation mode to Dossier, a collaborative rich-text editor used in multilingual, restrictive environments where public translation services may be unavailable or disallowed. The design lets users choose source and target languages, view a complete document in translation, copy text, and add comments while routing requests through the in-house Language Model Service. Engineers extracted shared translation components, added a Redux state slice and per-element cache, chunked content by paragraphs and other blocks, and used Intersection Observer, batching, debouncing, and caching to translate visible content incrementally; edits invalidate affected translations and trigger retranslation. To preserve sentence-level quality, inline formatting is stripped before translation, so the current translated view is plain text. The resulting mode is available in every Dossier, translating content lazily as users scroll and updating translated blocks when collaborators edit the document.


### [ClickHouse Cloud on Google Cloud Now Powered by Google Axion Processors: 30–55% Faster Queries, ~15% Fewer Compute Credits](https://yomu.fyi/post/clickhouse-cloud-on-google-cloud-now-powered-by-google-axion-processor.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Author: ClickHouse
- Published: Apr 22, 2026
- Type: Announcement

ClickHouse Cloud transitioned its Google Cloud infrastructure from x86 N2D instances to Arm-based Google Axion C4A instances to resolve CPU throttling and performance ceilings under peak query loads. The migration leverages Axion's dedicated physical cores per vCPU, higher memory bandwidth, local SSDs, and improved thermal and clock frequency stability. In automated ClickBench testing across 100 million rows, every query in the 43-query benchmark improved by 30% to 55%, while data load times decreased by approximately half. A production analysis of 38 migrated customer services demonstrated an average 15% reduction in compute credits consumed per query across diverse production workloads. ClickHouse Cloud executed this rollout using automated validation checks, scheduled maintenance windows, and capacity coordination with Google Cloud.


[Newer posts](https://yomu.fyi/topic/performance/page/20.md) · [Older posts](https://yomu.fyi/topic/performance/page/22.md)