What is an AI Readiness Speed Test?

An AI readiness speed test measures not only your download and upload speeds but also evaluates your connection's stability (jitter) and latency specifically for connecting to AI models like Claude, ChatGPT, and Gemini. It provides actionable metrics like Tokens Per Second (TPS) to ensure smooth token streaming and real-time agent responsiveness.

How are Provider TPS and Max Stream Rate calculated?

Provider TPS represents your connection's speed delivering real-time responses from specific LLM endpoints. It maps your Time to First Byte (TTFB) against a baseline rate, penalized by packet jitter. Max Stream Rate (Max TPS) simulates chunked token streams to measure latency standard deviation, calculating the absolute physical delivery stability limit of your connection, uncapped up to 1,500 TPS.

What does the Tokens Per Second (TPS) metric represent?

The Tokens Per Second (TPS) metric represents the maximum real-time stream capacity that your current internet connection supports for delivering AI tokens. While human-facing apps like ChatGPT's visual interface rendering in browsers are typically throttled or visually limited by browser rendering overhead, modern AI agents and programmatic pipelines consume raw streams directly. Flux Speed Test calculates the pure connection capability, showing that high-speed fiber with low jitter and low latency can easily sustain over 1,000+ TPS programmatically.

Why does Ping and Latency matter for Internet Speed?

While download speed determines how fast you can receive large files, ping (latency) measures the responsiveness of your connection. Low latency is critical for real-time applications such as competitive gaming, video conferencing, and quick AI query responses.

What is Bufferbloat and how does it affect my network?

Bufferbloat happens when your router buffers too much data, causing lag spikes when your network is under heavy load. A good bufferbloat grade means your connection remains stable and fast even when maxing out your broadband capacity.

How do I test my WiFi and Broadband performance accurately?

For the most accurate results, ensure you are close to your WiFi router or connected via Ethernet, pause any background downloads, and run our comprehensive speed and jitter test. We connect directly to the closest Cloudflare edge node to measure your true ISP (Internet Service Provider) capacity.

What is considered a good ping for competitive gaming?

For competitive multiplayer games (like shooters or fast-paced action), an ideal ping is under 20ms to 30ms. Anything below 50ms is generally considered good and lag-free. However, low jitter (consistency in latency) is just as important as a low ping to prevent lag spikes and rubber-banding.

How does a VPN affect my internet speed test?

A VPN (Virtual Private Network) routes your traffic through an intermediary server. This extra hop adds overhead and physical distance, which typically increases your latency (ping) and can reduce your maximum download/upload throughput. Flux Speed Test actively detects VPNs to provide context on why your connection might appear slower than expected.

How is the Video Streaming Quality score measured?

Our Video Stress module simulates loading 1080p and 4K Ultra HD video streams. It measures your connection's buffering risk by analyzing both sheer network speed (requiring >25 Mbps for stable 4K) and jitter consistency to predict if you will experience buffering or infinite loading wheels on platforms like Netflix or YouTube.

Why should I use Flux Speed Test over other tools?

Flux Speed Test uses 12 concurrent TCP streams routed through Cloudflare's global Anycast network to measure pure ISP capacity while minimizing the impact of single-server bottlenecks. We go beyond simple bandwidth tests by calculating RFC-compliant Jitter, actionable Bufferbloat grades, and an industry-first AI Readiness Score in TPS.

How can I contact the Flux Speed Test team?

For technical inquiries, support, ISP peering requests, or privacy inquiries, contact our team directly at info@fluxspeedtest.com.

INITIATE SPEED TEST

SERVER

SPEED RESULT

DOWNLOAD -- Mbps

UPLOAD -- Mbps

RESPONSIVENESS

Ping -- ms

Ud/Ld Latency -- / -- ms

Jitter (RFC) -- ms

Packet Loss 0%

Bufferbloat --

QUALITY SCORES

Gaming --/100

Video --/100

Web --/100

AI Readiness --/100

AI INSIGHTS

Claude (Anthropic) -- TPS

Gemini (Google) -- TPS

ChatGPT (OpenAI) -- TPS

Grok (xAI) -- TPS

WebSocket RTT -- ms

Max Stream Rate --

NETWORK DETAILS

ISP Frontier Fiber

Location Durham, NC

IP 192.168.1.42

RECENT TESTS IN YOUR AREA

Internet Speed Test Methodology - How We Measure Your Connection Speed

HOW WE MEASURE IT?

01. THROUGHPUT ARCHITECTURE

The Flux Network engine bypasses traditional HTTP latency by establishing 6 concurrent TCP streams directly to the Cloudflare Edge network.

Multi-Threading: We utilize the fetch API in parallel workers to saturate the available bandwidth, compensating for TCP slow-start protocols.
Cloudflare Anycast Routing: We do not rely on a single central server. Instead, our engine utilizes Cloudflare's global Anycast network. This technology automatically routes your connection to the physically closest of 330+ data centers worldwide. This is intentional: it ensures we measure your "Last Mile" connection speed (ISP capacity) rather than the latency of the open internet.
Server Verification: During the diagnostic process, we query the cdn-cgi/trace endpoint to cryptographically verify exactly which Colocation Center (Colo) you are connected to (e.g., IAD for Ashburn, LHR for London), providing total transparency on the test target.
Overhead Compensation: Results include a 1.05x multiplier to account for IP/TCP header overheads, providing a more accurate representation of Layer 1/2 capacity.

02. LATENCY & JITTER (RFC 3550)

Responsiveness is measured using high-frequency probes (~200ms interval) to the edge. We calculate Jitter using the standard Real-time Transport Protocol (RTP) formula defined in IETF RFC 3550.

J(i) = J(i-1) + (|D(i-1,i)| - J(i-1)) / 16

Where D(i-1,i) is the difference in packet arrival timing. This provides a exponentially weighted moving average that accurately reflects network stability for VoIP and Gaming.

03. BUFFERBLOAT & PACKET LOSS

Bufferbloat occurs when routers buffer too much data, causing high latency during transfers. We measure this by comparing Idle Latency vs. Loaded Latency (during max speed).

Grade A+

Delta < 5ms

Grade A

Delta < 30ms

Grade B

Delta < 60ms

Grade C

Delta < 150ms

Packet Loss is monitored via WebRTC Data Channels. If WebRTC is blocked, we use a heuristic based on failed HHR requests.

04. QUALITY OF EXPERIENCE (QoE)

Raw metrics are synthesized into application-specific scores (0-100).

Gaming Score Weighted blend: 50% Latency (<20ms ideal), 20% Jitter, 30% Throughput.
Video Score Evaluates buffering risk (Jitter) and HD capacity (Speed > 25Mbps).
Web Score Focuses on page load time simulation, rewarding low latency and moderate speed (>10Mbps).

05. AI READINESS

This module evaluates your connection's suitability for Large Language Models (LLMs) and real-time Generative AI streaming interfaces.

Provider Latency & Tokens/sec (TPS) We measure the Round Trip Time (RTT) of the initial SSL/TLS handshake (Time to First Byte / TTFB) to the edge gateways of major providers: Anthropic, Google, OpenAI, and xAI. We dynamically estimate real-time delivery performance based on network latency:
Provider TPS = Capped(800 * (60ms / TTFB) * (1 - Jitter / 100), 20, 1500)
Lower TTFB values scale up potential speed, while high packet jitter degrades delivery rates.
Max Stream Rate (Max TPS) We simulate streaming token bursts by fetching successive small payloads at high frequency. By calculating the standard deviation ($\sigma$) of chunk response times and factoring in raw bandwidth, we compute the connection's absolute physical streaming limits:
Max TPS = Capped(1200 - (σ * 10) + (DownloadSpeed * 0.2), 50, 1500)
This score rates the connection stability, identifying if background buffering or packet retransmissions will cause token generation stutter.
WebSocket RTT Measures the responsiveness of a persistent WebSocket connection, validating the low-overhead responsive channel required by conversational agents.
Privacy Shield VPN Detection Performs passive system timezone discrepancy checks and queries server-side ASN proxy identifiers. If a VPN/Proxy is detected, it triggers the split-view panel revealing the encryption provider.

Privacy Policy

Effective Date: September 2026 | Contact: info@fluxspeedtest.com

01. Commitment to Privacy

At Flux Speed Test, we believe network diagnostics should be fast, transparent, and respectful of your digital privacy. We operate on a strict privacy-first principle: we do not sell, rent, monetize, or track your personal browsing habits.

02. Technical Data We Process

To measure your broadband and AI readiness performance, our engine temporarily processes:

IP Address: Used solely to look up approximate city-level geolocation, ISP ASN, and route tests to the nearest Anycast edge node.
Performance Metrics: Download/upload throughput, idle/loaded latency, RFC 3550 jitter, bufferbloat grades, and AI model TTFB responsiveness.
Browser & Device Context: User-agent information used solely to calibrate web rendering performance scores.

03. How Your Data is Handled

Speed test results are processed in-flight. When you choose to share your results, an anonymized test record is saved with a unique UUID. IP addresses are masked and used solely for computing aggregated city-wide speed percentiles on our regional maps.

04. Zero Tracking Cookies

Flux Speed Test does not use intrusive third-party cross-site tracking cookies or marketing profiling pixels. Analytics on this site are privacy-friendly, cookieless, and GDPR/CCPA compliant.

05. Third-Party Diagnostic Infrastructure

To deliver accurate high-performance tests, we interact with trusted global infrastructure:

Cloudflare Anycast Network: For low-latency multi-stream TCP throughput benchmarking.
Carto & OpenStreetMap: For rendering regional speed distribution map boundaries.

06. Contact & Data Rights

You have the right to request deletion of any shared test record associated with your unique test ID. For any privacy questions or data requests, contact our privacy officer at info@fluxspeedtest.com.

TCP THROUGHPUT

LATENCY / GAMING

VIDEO STRESS

AI READINESS

METHODOLOGY

01. THROUGHPUT ARCHITECTURE

02. LATENCY & JITTER

03. BUFFERBLOAT

04. SCORING

05. AI READINESS

Free Internet Speed Test - Check Download Speed, Upload Speed, Ping, Latency & AI Readiness

TCP THROUGHPUT

LATENCY / GAMING

VIDEO STRESS

AI READINESS

Internet Speed Test Methodology - How We Measure Your Connection Speed

HOW WE MEASURE IT?

01. THROUGHPUT ARCHITECTURE

02. LATENCY & JITTER (RFC 3550)

03. BUFFERBLOAT & PACKET LOSS

04. QUALITY OF EXPERIENCE (QoE)

05. AI READINESS

Frequently Asked Questions

What is an AI Readiness Speed Test?

How are Provider TPS and Max Stream Rate calculated?

What does the Tokens Per Second (TPS) metric represent?

Why does Ping and Latency matter for Internet Speed?

What is Bufferbloat and how does it affect my network?

How do I test my WiFi and Broadband performance accurately?

What is considered a good ping for competitive gaming?

How does a VPN affect my internet speed test?

How is the Video Streaming Quality score measured?

Why should I use Flux Speed Test over other tools?

How can I contact the Flux Speed Test team?

Privacy Policy

01. Commitment to Privacy

02. Technical Data We Process

03. How Your Data is Handled

04. Zero Tracking Cookies

05. Third-Party Diagnostic Infrastructure

06. Contact & Data Rights

METHODOLOGY

01. THROUGHPUT ARCHITECTURE

02. LATENCY & JITTER

03. BUFFERBLOAT

04. SCORING

05. AI READINESS