--- title: Benchmarks | Telnyx canonical: https://telnyx.com/benchmarks --- # Benchmarks | Telnyx Benchmarks # Inference & Voice AI Benchmarks Latency, throughput and reliability for Telnyx Inference, Voice AI, Speech and Web Search, measured against the providers you would otherwise be using. Same prompts, same profiles, same day. We publish the runs we lose. [See the results](#results)[Methodology](#method)[All reports](#reports) Latest inference run15 Oct 2026 ## GLM 5.3 Flash, median end to end latency Pooled across four workloads at concurrency 1 and 10\. Fresh prompts, streaming, reasoning effort low. Lower is better. Telnyx 4.92s Modal Shared 8.75s 240 of 240 requests completed on both providers. Modal Shared had the lower p95 time to first visible answer in seven of eight cells. Covers Modal Shared only, not Modal Dedicated or Stateful Actors. [Full report and all eight cells ↗](https://telnyx.com/resources/glm53-flash-benchmark) 14 Published benchmark reports 4 Products under measurement p95 Reported on every latency, not just the median 3 of 14 Reports where a rival wins the headline metric Results ## Benchmarks by product. Pick a product. Every result set below names its metric, its run date, its workload profile and the number of attempts behind it. Bold marks the best value in a column, whoever it belongs to. Product Inference7Voice AI4Speech6Web Search1 7 result sets · latest run 15 Oct 2026 · identical prompts across providers, plus multi-turn and one Telnyx-only model comparison InferenceTelnyx run15 Oct 2026 ### GLM 5.3 Flash on Telnyx Inference against Modal Shared Metric end to end latency, lower is better, and output speed, higher is better. Model zai-org/GLM-5.3-Flash on both, reasoning effort low, streaming. Profiles four, not six: about 1k, 10k and 100k input with 256 output, and 10k input with 1,024 output, each at concurrency 1 and 10. Attempts 240 per provider, fresh prompts, 30 per cell after three discarded warm ups. All 240 completed on both. Pricing published token rates 70 percent lower on Telnyx: $0.135 input and $0.450 output per 1M, against $0.450 and $1.500. Scope Modal Shared only, not Modal Dedicated or Stateful Actors. __GLM 5.3 Flash on Telnyx Inference against Modal Shared__ | Provider | End to end p50seconds, lower better | End to end p95seconds, lower better | Output speed p50tokens per second, higher better | Successful requestsof 240 | | ------------ | ----------------------------------- | ----------------------------------- | ------------------------------------------------ | ------------------------- | | Telnyx | 4.92 | 11.91 | 138.1 | 240 | | Modal Shared | 8.75 | 20.62 | 66.2 | 240 | Telnyx finished sooner and generated about 2.1 times faster. Modal Shared had the lower p95 time to first visible answer in seven of eight cells, so a chat UI that optimises for the first words should weigh that column too. [Full report ↗](https://telnyx.com/resources/glm53-flash-benchmark) InferenceTelnyx run08 Oct 2026 ### Multi-turn task completion across three Telnyx-hosted models Comparison Telnyx only. Three models on the same Telnyx Inference API, no other provider in this table. Scenarios 15 four-turn conversations per model across developer, business, planning, editing and structured output workflows. Scoring two blind judges, DeepSeek V4.1 Flash and MiniMax M3, for 30 completion decisions per model, plus 28 deterministic checks on the final answers. Settings reasoning effort low, temperature 0, 4,096 token ceiling. Reliability all three models completed every request with no HTTP errors, timeouts or truncated responses. __Multi-turn task completion across three Telnyx-hosted models__ | Model | Judge completionof 30, higher better | Objective checksof 28, higher better | Conversation p50seconds, lower better | Conversation p95seconds, lower better | | ------------- | --------------------------------------- | ------------------------------------ | ------------------------------------- | ------------------------------------- | | Kimi K3 | 305.00 of 5 on corrections, both judges | 23 | 9.36 | 55.71 | | GLM 5.3 | 27 | 23 | 16.38 | 53.19 | | GLM 5.3 Flash | 27 | 22 | 8.35 | 31.27 | Kimi K3 was strongest on completion and correction following. GLM 5.3 Flash was fastest at median and p95 total conversation time. GLM 5.3 matched Kimi K3 on objective checks, trading latency for depth. [Full report ↗](https://telnyx.com/resources/choose-open-weight-model) InferenceTelnyx run07 Oct 2026 ### Multi-turn conversations across three models and four providers Comparison Telnyx, Baseten, Together AI and Fireworks, all called through one orchestrator with explicit conversation history in every request. Scenarios 15 fixed multi-turn workflows across developer, business and structured output tasks, three repeats each: 45 attempted conversations per provider-model pair, 540 in total. Metric total conversation response time, the sum of model response time across the workflow, excluding simulated user pauses. Settings reasoning effort low, temperature 0, 4,096 token ceiling per turn. Scoring one blind judge, DeepSeek V4.1 Flash, plus deterministic checks. Mean blind quality scores ranged from 4.79 to 5.00 with a different leader per model family. Reliability of 135 attempted conversations per provider: Telnyx and Baseten completed 134, Fireworks 133, Together AI 129\. Failed conversations stay excluded from the latency percentiles. __Multi-turn conversations across three models and four providers__ | Model | Telnyx p50seconds conversation, lower better | Next fastest p50seconds conversation, lower better | Telnyx observed p95seconds conversation, lower better | Next fastest observed p95seconds conversation, lower better | | ------------- | -------------------------------------------- | -------------------------------------------------- | ----------------------------------------------------- | ----------------------------------------------------------- | | GLM 5.3 Flash | 8.26 | 11.0025 percent higher | 35.62 | 42.7217 percent higher | | GLM 5.3 | 13.48 | 15.8515 percent higher | 41.33 | 50.65Baseten. Together AI observed 181.20, Fireworks 118.83 | | Kimi K3 | 10.49 | 15.9534 percent higher | 46.40 | 49.667 percent higher | Telnyx recorded the lowest median and observed p95 total conversation time for all three model families. The other metrics are separately read, not averaged: quality was tightly grouped and had a different leader per model (Together AI on GLM 5.3 Flash, Telnyx on GLM 5.3, Baseten on Kimi K3), Together AI had the fastest median first visible answer on both GLM models while Telnyx led it on Kimi K3, and reliability varied by provider-model pair. Observed p95 values describe this sample and are not SLAs. [Full report ↗](https://telnyx.com/resources/multi-turn-llm-benchmark) InferenceTelnyx run18 Sep 2026 ### GLM 5.3 end to end latency across four providers Metric end to end, streaming, lower is better. Profile shown 88.7k input / 1,200 output. Prompts 60 distinct, identical across providers. Settings temperature 0, reasoning effort high. Attempts 240, of which 230 completed. Telnyx was called from a stateful actor in us-east-2, the other three over the public internet. __GLM 5.3 end to end latency across four providers__ | Provider | End to end p50seconds, lower better | End to end p95seconds, lower better | | ------------ | ----------------------------------- | ----------------------------------- | | Telnyx | 9.87 | 12.80 | | Together AI | 14.67 | 69.64 | | Fireworks AI | 20.61 | 26.84 | | Baseten | 21.81 | 23.18 | Lowest median in five of six profiles and the lowest p95 in all six. The p95 column is the one to read: a 69.64 second tail is shared capacity under load, not a slow model. [Full report ↗](https://telnyx.com/resources/glm-5-3-latency-benchmarks-provider) InferenceTelnyx run02 Sep 2026 ### GLM 5.3 Flash throughput and completion rate Metric output throughput, higher is better, and how many of 60 runs returned at all. Settings streaming, temperature 0\. Completion rate is reported because a provider that refuses one run in four is not comparable on speed alone. __GLM 5.3 Flash throughput and completion rate__ | Provider | Throughput p50tokens per second, higher better | Runs completedof 60 | | ------------ | ---------------------------------------------- | ------------------- | | Telnyx | 196.4 | 60 | | Baseten | 190.116 runs returned HTTP 429, rate limited | 44 | | Fireworks AI | 127.9 | 60 | | Together AI | 102.4 | 60 | Baseten is within 3 percent of us on the runs that completed. It is the 16 that did not that decide whether you can build on it. [Full report ↗](https://telnyx.com/resources/glm-5-3-flash-benchmarks) InferenceTelnyx run25 Jun 2026 ### GLM 5.2, where first token and finished answer disagree Metric both, lower is better. Attempts 240 completed. Note Telnyx served FP8, the others served the unquantized weights; this is disclosed because it affects both columns. Reasoning was enabled, so the TTFT column records the first token of any kind, which may be reasoning rather than answer text. TTFA is not in this run. __GLM 5.2, where first token and finished answer disagree__ | Provider | TTFT p50seconds, lower better, reasoning included | End to end p50seconds, lower better | | ------------ | ------------------------------------------------- | ----------------------------------- | | Telnyx | 1.12zai-org/GLM-5.2-FP8 | 6.00 | | Baseten | 0.76 | 6.19 | | Together AI | 0.79 | 7.49 | | Fireworks AI | 1.28 | 14.98 | Baseten shows first activity 360ms sooner and finishes 190ms later. Until TTFA is measured, neither column tells you when the answer itself starts appearing. [Full report ↗](https://telnyx.com/resources/inference-benchmark-ttft-vs-e2e) InferenceTelnyx run28 Jul 2026 ### The same workload, called from inside the region Metric both, lower is better. What changed only the caller. Identical model, identical prompts, moved from the public internet to a Telnyx stateful actor colocated with the GPUs. Caveat, stated in the source the comparison uses the earlier provider baselines rather than fresh same day reruns. __The same workload, called from inside the region__ | Configuration | TTFT p50seconds, lower better, reasoning included | End to end p50seconds, lower better | | -------------------------------------------------- | ------------------------------------------------- | ----------------------------------- | | GLM 5.2, called over the public internet | 1.12 | 6.00 | | GLM 5.2, called from a stateful actor in region | 0.76 | 4.54 | | MiniMax M3, called over the public internet | 1.87 | 7.56 | | MiniMax M3, called from a stateful actor in region | 0.61 | 7.27 | TTFT falls 32.5 percent on GLM 5.2 and 67.3 percent on MiniMax M3\. Nothing about the model changed. The distance did. [Full report ↗](https://telnyx.com/resources/reducing-latency-stateful-actors) What the numbers say ## Why the numbers look like this. Four claims. Each one is only here because a table above pays for it, and each links to the run it came from. 01 ### We own the infrastructure. Our GPUs sit on our own network, not rented capacity resold to whoever bid. You cannot see that in a median. You see it in the tail, when everybody else's tenants show up at once. 12.80s observed p95 against 69.64s GLM 5.3, 88.7k input tokens. Telnyx recorded a 12.80s observed p95, compared with 69.64s for Together AI in this sample. [The run ↗](https://telnyx.com/resources/glm-5-3-latency-benchmarks-provider) 02 ### It runs where your workload runs. Compute in region is not a data residency checkbox. It is distance, and distance is most of what a first token costs. Move the caller next to the GPU and nothing else has to change. TTFT down 67.3 percent MiniMax M3, 1.87s to 0.61s. Same model, same prompts, called from a stateful actor beside the GPUs instead of across the internet. [The run ↗](https://telnyx.com/resources/reducing-latency-stateful-actors) 03 ### The whole agent sits on one network. Telephony, speech to text, the model and text to speech on separate vendors means four networks between a caller and an answer. We measure the stack, because the stack is what the caller hears. Above the Pareto frontier ServiceNow Research, 150 live conversations. The first hosted voice API on the leaderboard to beat the Pipecat agents on quality at equal accuracy. 04 ### We publish the runs we lose. A vendor benchmark that a vendor always wins tells you the vendor chose the workload. Three of the fourteen reports on this page have somebody else at the top of the headline column, and all three are still here. Brave 561ms, Telnyx 1,027ms Web search latency, 20 identical queries. Baseten also shows first activity 360ms sooner than us on GLM 5.2. [The run ↗](https://telnyx.com/resources/web-search-api-benchmark) Methodology ## How every run is made. The same six rules govern every table on this page. Where a run breaks one of them, the table says so in its own header rather than in a footnote nobody reads. 01 1. Identical prompts, sent the same day Every provider in a table gets the same prompt set inside the same session. A number carried over from an earlier run is labeled as a baseline, not presented as a comparison. 02 2. Six workload profiles, never one 1k, 10k and roughly 100k input tokens, against short and long outputs. A single profile is how a provider picks the shape of the workload it happens to be good at. 03 3. Distributions, not averages p50 and p95 together, always. A median tells you what usually happens. p95 tells you what your support queue will look like. 04 4. Failed calls are counted, not dropped A rate limited request is a result. Where a provider returned fewer runs than it was asked for, the table shows the completion rate beside the speed, because excluding the failures flatters whoever failed. 05 5. Asymmetries are disclosed in the table Quantization, model identifier, and where the request originated from. Our GLM 5.2 run served FP8 against unquantized weights elsewhere, and our traffic came from a stateful actor in region. Both are stated where the numbers are, not at the bottom of the page. 06 6. The runs we lose stay published A result is not withdrawn because a competitor came out ahead. Three of the reports above have somebody else at the top of the headline column. Check every number at its source Each result we ran links to the full report behind it. Our test harness is internal, so the published reports are the record. To see how your own traffic compares, send your prompts to the same endpoints from a free account. [Browse the report archive](#reports) * Run date and settings sit above every result set * Superseded runs stay in the archive below * Results a rival wins stay published Archive ## Every report we have published. Newest first, including the runs that now sit behind a newer one. Nothing is removed when it is superseded. * [20 Oct 2026InferenceMulti-turn LLM benchmark across three models and four providersLowest median and observed p95 conversation time, all three models Quality, reliability and first answer read separately](https://telnyx.com/resources/multi-turn-llm-benchmark) * [15 Oct 2026InferenceGLM 5.3 Flash benchmark: Telnyx against Modal Shared4.92s median end to end Modal leads on p95 first answer](https://telnyx.com/resources/glm53-flash-benchmark) * [13 Oct 2026SpeechTelnyx STT against ElevenLabs Scribe on real phone audio1.0s median to first words, live Scribe leads on accuracy](https://telnyx.com/resources/telnyx-stt-vs-elevenlabs-scribe-benchmark) * [08 Oct 2026InferenceThree open-weight models tested on multi-turn agentsKimi K3 30 of 30 completed GLM 5.3 Flash fastest](https://telnyx.com/resources/choose-open-weight-model) * [18 Sep 2026InferenceGLM 5.3 latency benchmarks across four inference providers9.87s median end to end Lowest p95 in all six profiles](https://telnyx.com/resources/glm-5-3-latency-benchmarks-provider) * [17 Sep 2026Web SearchBenchmarking Tavily, Telnyx, DuckDuckGo and Brave for agent workflows1,027ms mean latency Second of four, Brave wins](https://telnyx.com/resources/web-search-api-benchmark) * [04 Sep 2026Voice AIVoice AI agents compared on latency71ms carrier leg p50 Aggregates four third party studies](https://telnyx.com/resources/voice-ai-agents-compared-latency) * [03 Sep 2026InferenceGLM 5.3 Flash inference benchmarks on Telnyx196.4 tokens per second p50 60 of 60 runs completed](https://telnyx.com/resources/glm-5-3-flash-benchmarks) * [01 Sep 2026InferenceDeepSeek V4 Flash inference benchmarks on Telnyx12.24s p95 end to end Against 14.49s next best](https://telnyx.com/resources/deepseek-v4-flash-benchmarks) * [28 Jul 2026InferenceRunning repeatable LLM latency benchmarks with stateful actorsTTFT down 67.3 percent Same model, called in region](https://telnyx.com/resources/reducing-latency-stateful-actors) * [25 Jun 2026InferenceGLM 5.2 inference benchmarks6.00s end to end p50 Third of four on first token](https://telnyx.com/resources/glm-5-2-inference-benchmarks-provider) * [16 Jun 2026InferenceMiniMax M3 runs best on Telnyx Inference7.56s end to end p50 14.72s p95 against 54.06s](https://telnyx.com/resources/inference-minimax-m3-benchmark) * [28 Apr 2026InferenceWhich latency metric should you optimize for540 streamed completions The measurement key on this page](https://telnyx.com/resources/inference-benchmark-ttft-vs-e2e) * [10 Dec 2025SpeechEarly benchmarks on ElevenLabs against MiniMax text to speechQualitative read, three scenarios Superseded by the current harness](https://telnyx.com/resources/minimax-speech-2-6-vs-elevenlabs-tts-benchmarks) Start building ## Run it against your own workload. Our profiles are a guess at your traffic. Yours are not. Free tier, no card, and the same infrastructure every number on this page was measured on. [Sign up](https://portal.telnyx.com/#/login/sign-up)[Inference](https://telnyx.com/products/inference)[Voice AI agents](https://telnyx.com/products/voice-ai-agents)[Talk to sales](https://telnyx.com/contact-us) [Start building](https://telnyx.com/sign-up) ### Ask AI * [![GPT](https://d2efw2akbx8gaf.cloudfront.net/https://dotcom-payload-assets-prod.s3.us-east-1.amazonaws.com/cms-media/chatgpt_icon-ea55e09d.jpeg?width=92&format=webp)](https://chat.openai.com/?q=What makes Telnyx's global network different, breakdown call quality for normal phone calls and voice AI calls.) * [![Claude](https://d2efw2akbx8gaf.cloudfront.net/https://dotcom-payload-assets-prod.s3.us-east-1.amazonaws.com/cms-media/claude_icon-c7c8eef3.webp?width=92&format=webp)](https://claude.ai/new?q=What makes Telnyx's global network different, breakdown call quality for normal phone calls and voice AI calls.) * [![Perplexity](https://d2efw2akbx8gaf.cloudfront.net/https://dotcom-payload-assets-prod.s3.us-east-1.amazonaws.com/cms-media/perplexity_icon-cd5b74f1.webp?width=92&format=webp)](https://www.perplexity.ai/search?q=What makes Telnyx's global network different, breakdown call quality for normal phone calls and voice AI calls.) * [![Gemini](https://d2efw2akbx8gaf.cloudfront.net/https://dotcom-payload-assets-prod.s3.us-east-1.amazonaws.com/cms-media/gemini_icon-4ff1cbf4.webp?width=92&format=webp)](https://www.google.com/search?udm=50&aep=11&q=What makes Telnyx's global network different, breakdown call quality for normal phone calls and voice AI calls.) * [![Grok](https://d2efw2akbx8gaf.cloudfront.net/https://dotcom-payload-assets-prod.s3.us-east-1.amazonaws.com/cms-media/grok_icon-e588e5bd.webp?width=92&format=webp)](https://x.com/i/grok?text=What makes Telnyx's global network different, breakdown call quality for normal phone calls and voice AI calls.)