--- title: "Phonely Alma Review: Voice LLM Cost & Latency" canonical: https://wavect.io/blog/phonely-alma-voice-llm-review/ language: en description: "Phonely Alma review: verify its 182 ms TTFT, 206 ms P99, $0.55 blended token cost, voice-agent limits and production pilot criteria." image: "https://wavect.io/img/blog/headers/header_phonely-alma-voice-llm-review.png" --- [**Back**](/blog/overview/) [![Kevin Riedl](/img/team/kevin.webp)](/team/kevin-riedl/) [Kevin Riedl](/team/kevin-riedl/) https://linkedin.com/in/wsdt 11 min read · 6 Sep 2026 Last reviewed September 6, 2026 [**Next**](/blog/agent-knowledge-transfer-cheaper-models/) # Phonely Alma Review: Is the Voice LLM Ready for Production? TL;DR Phonely's Alma is a specialist language model for structured phone conversations, not a general coding or long-horizon agent model. Phonely reports 182 ms time to first token, 206 ms P99, 379 ms full replies, a 0.867 phone-etiquette score and $0.55 per blended million tokens in a 200-transcript benchmark run on 21 August 2026. Those results make Alma a credible pilot candidate for scheduling, support, lead qualification and other repeatable call flows. They do not establish independent superiority or end-to-end caller latency because the vendor ran and graded the test, used sequential requests, and did not publish every prompt, transcript, judge, region or concurrency condition. Compare Alma with the incumbent on real calls, including accents, noise, interruptions, tool calls, policy adherence, escalation, P95 and P99 voice-to-voice latency, cost per successful call, data terms and fallback behavior. Adopt it only when the complete call outcome improves. **Alma is a credible specialist LLM for a narrow job: keeping structured phone conversations fast, concise and on task.** Phonely reports 182 ms to the first token, 206 ms at P99, 379 ms for a full reply and $0.55 per blended million tokens. Those numbers earn Alma a production pilot. They do not prove that it is the best voice model for your callers. This review owns the model-selection decision inside a voice-agent stack. For the adjacent components, use our separate [streaming speech-to-text production review](/blog/nvidia-nemotron-3-5-asr-production-review/) and [self-hosted versus API text-to-speech analysis](/blog/miso-tts-self-hosted-vs-api/). Keeping the three decisions separate prevents a fast LLM benchmark from being mistaken for a fast call. ## What is Phonely Alma? **Alma is a language model optimized for repeatable phone workflows, conversational phrasing and tool calling.** Phonely positions it for appointment booking, customer support, lead qualification, debt collection, lookups, transfers and other flows with a definable outcome. The company explicitly says it is not designed for coding, creative writing, open-ended multi-agent work or long-horizon planning. The term “voice-native LLM” needs care. Alma is priced in input and output tokens, and Phonely says it can work with existing transcription and speech-generation components. Evaluate it as the language-model stage unless your chosen deployment contract explicitly includes audio input, audio output, turn detection and telephony. A specialist text LLM can still be excellent for voice. It just does not remove the other clocks in the pipeline. ## What does Phonely claim for Alma? | Metric | Alma | GPT-4.1 | GPT-5.6 | Buyer interpretation | | --- | --- | --- | --- | --- | | Time to first token | 182 ms | 490 ms | 997 ms | Useful LLM-stage responsiveness signal | | P99 response latency | 206 ms | 2.02 s | 2.84 s | Promising tail result under the published test | | Full reply | 379 ms | 771 ms | 1.46 s | Suggests short spoken answers are an optimization target | | Phone etiquette | 0.867 | 0.700 | 0.770 | Directionally useful, but rubric and judges matter | | Blended cost per 1M tokens | $0.55 | $3.50 | Not priced | Model-token comparison, not total call cost | The [official Alma launch page](https://www.phonely.ai/alma) says the comparison used 200 held-out call transcripts, identical prompts and grading, sequential requests from an AWS `us-east-1` client, and five excluded warm-ups. Phonely also publishes shared pricing of $0.30 per million input tokens and $1.30 per million output tokens, priority pricing of $0.50 and $2.10, plus custom VPC or on-premises terms. Independence and trademarks Wavect publishes this page and is itself a provider, so we have a commercial interest in it. We are not affiliated with, endorsed by or partnered with the other companies named here, and all third-party company names, brands and trademarks are the property of their respective owners. Statements about other providers are taken from publicly available sources, primarily their own published pages, as of the review date shown on this page, and may have changed since. Please verify them directly before you decide. This page was written to the best of our knowledge and with the intent to remain objective. If you believe anything here is inaccurate or unfair, write to us and we will correct it: [office@wavect.io](mailto:office@wavect.io) ## What does the Alma benchmark prove? **It proves that Alma deserves a head-to-head test for structured phone work.** The vendor names the models, date, client region, sample size and several latency measures. Publishing a P99 is especially useful because one long silence can damage a call more than a strong median can rescue it. It does not establish general superiority. The evaluation is vendor-run and appears to use transcripts rather than complete live audio calls. The public page does not expose the full held-out set, prompt, phone-etiquette rubric, judge composition, confidence intervals, request lengths, cache state, rate tiers or concurrent-load profile. Sequential requests do not tell a buyer what happens at 50, 500 or 5,000 simultaneous calls. The OpenAI and Phonely entries are also products with different scopes. A specialist can beat a larger general model on one constrained distribution without being a better general model. That is the product thesis, not a flaw. The safe claim is “faster and cheaper in Phonely’s published phone benchmark,” not “faster, cheaper and more accurate for every voice agent.” ## Why can a specialist LLM work better on phone calls? Phone dialogue punishes behaviors that look harmless in chat. Long preambles delay the useful answer. Repeating the user sounds robotic. A missing confirmation can corrupt an address, date or payment promise. Tool calls must match what the agent says. Interruptions, fragments, spelling, numbers and background noise change the input distribution. A specialist model can trade broad knowledge and long reasoning for four valuable properties: - **Short answer starts:** it can begin with the next useful phrase instead of a written-style introduction. - **Conversation discipline:** it can confirm ambiguous names, emails and dates without producing a lecture. - **Flow adherence:** it can collect required fields, call the correct tool and keep spoken claims aligned with system state. - **Predictable serving:** a smaller, controlled model can reduce tail latency and make reserved capacity economical. This is also why generic academic scores are insufficient. The [open PhoneBench evaluation design](https://www.pipecat.ai/benchmarks/phonebench-alpha-1) measures telephone speaking style, tool-call accuracy, say-do consistency, factual grounding, coherence, authentication, escalation and caller outcome together with latency and estimated cost per minute. That is much closer to a procurement scorecard than a single general reasoning score. ## Why 182 ms is not the caller’s wait time **Time to first token measures one component, not voice-to-voice latency.** A cascaded agent still needs telephony ingress, jitter handling, speech-to-text, end-of-turn detection, context or knowledge retrieval, the LLM, tool execution, text-to-speech and audio delivery. Some work overlaps, but the user experiences the complete chain. Measure at least these clocks: 1. last audible user sample to stable transcript; 2. stable transcript to first useful LLM token; 3. first LLM token to first audible synthesized speech; 4. last user sample to first audible agent speech at P50, P95 and P99; 5. tool-request start to confirmed business-state change; 6. interruption to agent stop, plus false interruption rate. A recent research benchmark reinforces the need for complete-call evaluation. The [τ-Voice study](https://arxiv.org/abs/2603.13686) combines grounded task completion, full-duplex interaction and realistic audio across 278 tasks. Under its setup, tested voice agents retained only 30 to 45 percent of the text baseline’s task capability in realistic conditions, with accents, noise and turn-taking exposing large gaps. Fast first tokens do not repair a wrong tool call. ## How much can Alma actually save? The published $0.55 blended rate is about 84 percent below the $3.50 GPT-4.1 comparator. If one call consumes 8,000 blended LLM tokens, the illustrative model charge is about $0.0044 for Alma versus $0.028 for the comparator, a difference of $0.0236 per call. At 100,000 calls, that is $2,360 before volume discounts. This arithmetic is not a quote. “Blended” hides the input-output mix, and the total cost still includes telephony, transcription, speech generation, retrieval, orchestration, storage, observability, support and human escalations. Priority capacity also has higher input and output rates than shared capacity. Use this formula: `cost per resolved call = all platform and model charges + allocated operations + human handling / calls that reach the accepted outcome` A cheaper model that transfers more callers, repeats steps or books the wrong slot can cost more per resolution. A specialist that shortens answers may save both tokens and call minutes, but only production traces can establish that benefit. ## Shared, priority or on-premises Alma? | Path | Best first fit | Main question before purchase | | --- | --- | --- | | Shared API | Low-risk evaluation and variable early traffic | What P95 and P99 apply under your region, prompt size and bursts? | | Priority API | Production calls with a latency objective | What capacity, support and service commitments are contractual? | | VPC or on-premises | Sensitive data, residency or network-control requirements | Who operates upgrades, scaling, failover, monitoring and incident response? | Do not equate a deployment label with compliance. Ask for the data-processing agreement, subprocessors, regions, retention defaults, deletion behavior, training-use terms, redaction boundary, encryption, tenant isolation, audit logs, incident notification, model update policy and exit path. For on-premises, identify whether audio, transcripts, prompts, telemetry or support bundles ever leave your environment. ## A production pilot scorecard for Alma | Dimension | Minimum metric | Failure that the average hides | | --- | --- | --- | | Task outcome | Correct resolution rate by call type | Easy FAQ calls masking failed bookings | | Tool use | Correct tool, arguments and state change | Agent says “booked” after a failed write | | Conversation | Human-rated brevity, clarity and repair | Natural voice reading an awkward paragraph | | Latency | Voice-to-voice P50, P95 and P99 | One percent of turns stalling for seconds | | Interruptions | Stop latency and false barge-in rate | Agent talks over short pauses or background speech | | Safety and policy | Forbidden-action and required-disclosure rate | Fast completion that skips authentication | | Economics | Cost per accepted resolution | Cheap tokens with more transfers and retries | | Reliability | Error rate, timeout rate and fallback success | Clean demo with no burst or provider failure | ## How to test Alma against your current LLM 1. **Choose one bounded flow.** Appointment changes or lead qualification are better pilots than “answer every customer question.” 2. **Freeze 200 to 500 representative scenarios.** Include real distributions, rare high-cost cases, accents, noise, numbers, spelling, interruptions, silence and hostile inputs. 3. **Keep the stack constant.** Use the same STT, TTS, tools, knowledge, prompt facts, region and telephony path. Change the LLM first. 4. **Blind the review.** Human reviewers should score outcomes and conversation quality without seeing the model name. 5. **Load-test the paid tier.** Reproduce expected concurrency and bursts. Report P50, P95 and P99 rather than a single best run. 6. **Force business failures.** Reject a calendar write, delay a CRM, return stale knowledge, drop an API response and verify truthful recovery. 7. **Run shadow traffic.** Replay consented, appropriately protected calls or scenarios before exposing the candidate to live callers. 8. **Release reversibly.** Route a small call slice, retain a tested fallback and define rollback thresholds for outcome, safety, latency and cost. ## Who should pilot Alma now? Pilot Alma if your call flows are repeatable, LLM latency or cost is material, concise phone behavior matters, tools have clear success states, and your team can compare production outcomes. Scheduling, qualification, routing and structured support fit that profile. Keep a general model or hybrid route for open-ended research, long planning, creative work, broad troubleshooting and calls whose next step cannot be defined. A model router can let Alma own the live conversation while a different model handles a slower back-office task. Our [LLM gateway and routing guide](/blog/llm-gateway-router-comparison-2026/) covers that portability decision. ## Sources and verification boundary Alma capabilities, use cases, deployment choices, prices and benchmark figures come from Phonely’s official launch page. The buyer-side scorecard is informed by PhoneBench’s public evaluation dimensions and the τ-Voice research design. Facts were checked on 6 September 2026. Wavect did not run Alma, inspect its weights or training corpus, verify the reported ten million calls, or reproduce Phonely’s latency, etiquette, uptime, accuracy, cost or carbon claims. Treat every unreplicated result as vendor-reported. ## Frequently Asked Questions About Phonely Alma ### What is Phonely Alma? Alma is a specialist language model built for structured phone conversations. Phonely positions it for scheduling, support, qualification, collections, tool calls and concise conversational behavior, not coding or long-horizon general agent work. ### Is Alma a speech-to-speech model? Do not assume that from the voice-native label. Phonely prices Alma in text tokens and says it works with existing transcription and text-to-speech components. Evaluate the contracted product boundary and measure the whole audio pipeline. ### Is Alma really faster than GPT-4.1? In Phonely’s vendor-run test of 200 held-out call transcripts, Alma reported 182 ms time to first token versus 490 ms for GPT-4.1, plus 206 ms versus 2.02 seconds at P99. Reproduce the result under your prompts, region, tier, concurrency and full call stack. ### How much does Alma cost? On 6 September 2026, Phonely listed shared rates of $0.30 per million input tokens and $1.30 per million output tokens, and priority rates of $0.50 and $2.10. It reported $0.55 per blended million tokens in its benchmark. VPC and on-premises pricing is custom. ### What should an Alma pilot measure? Measure correct call outcomes, tool-state accuracy, policy adherence, human-rated conversation quality, interruption behavior, voice-to-voice P50, P95 and P99, failures, fallback success and total cost per accepted resolution. ### Should every voice agent switch to Alma? No. Alma fits repeatable, definable phone flows. Keep or compare a general model for open-ended reasoning and long tasks. Switch only after a controlled test shows better complete-call outcomes without new safety, reliability or data risks. ## Final thoughts Alma makes a strong case for specialization. Phonely’s published 182 ms TTFT, 206 ms P99 and $0.55 blended token rate are relevant to voice teams, and the company is refreshingly clear that this is not a general-purpose model. The evidence supports a pilot, not a universal winner. A buyer still needs complete audio latency, task success, tool-state consistency, interruption quality, policy adherence, concurrent-load behavior, data terms and cost per resolved call. Keep STT, LLM and TTS decisions modular. Test one bounded flow against the incumbent, then route production traffic only when the whole call improves. Choose the model that resolves the call correctly, quickly and economically, not the one with the smallest isolated number. ## You may also like.. [**Evaluate the speech-to-text stage** Separate stable transcript and end-of-turn latency from the LLM benchmark with a production review of streaming ASR.](/blog/nvidia-nemotron-3-5-asr-production-review/) [**Evaluate the text-to-speech stage** Compare managed and self-hosted speech generation by latency, privacy, hardware and total operating cost.](/blog/miso-tts-self-hosted-vs-api/) Models and infrastructure ## Continue through this cluster Model selection, inference economics, local deployment, compression and serving architecture. [Start with the cornerstone**Self-Hosting LLMs in the EU: When Open Weights Actually Pay Off**](/blog/self-hosting-llms-eu-cost/) - [NVIDIA PAIR Review: Local AI Routing and the AMD Gap](/blog/nvidia-pair-amd-rocm-strix-halo/) - [Tencent Hy4 Preview Review: Is the 1M-Context Coding Model Worth a Pilot?](/blog/tencent-hy4-preview-coding-agent-review/) - [PageLM Review: Self-Hosting and Commercial Use](/blog/pagelm-self-hosted-ai-study-platform/) - [M6 Mac mini vs M5 Mac Studio for Local AI: Buyer's Guide](/blog/mac-mini-m6-vs-mac-studio-m5-local-ai/) - [FreeToken AI Review: Run Frontier MoE Models on Consumer GPUs?](/blog/freetoken-ai-inference-engine-review/) [**Back**](/blog/overview/) [![Kevin Riedl](/img/team/kevin.webp)](/team/kevin-riedl/) [Kevin Riedl](/team/kevin-riedl/) https://linkedin.com/in/wsdt 11 min read · 6 Sep 2026 Last reviewed September 6, 2026 [**Next**](/blog/agent-knowledge-transfer-cheaper-models/) ## Structured Data ```json { "@context": "https://schema.org", "@graph": [ { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ], "employee": [ { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } } ], "founder": [ { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } } ], "legalRepresentative": [ { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } } ], "name": "Wavect GmbH", "subjectOf": { "@id": "https://wavect.io/verified-claims.json#dataset", "@type": "Dataset", "creator": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] }, "description": "A machine-readable registry of quantitative and qualitative claims published by Wavect, with review dates, localized page appearances and public third-party citations where available.", "inLanguage": "en", "isAccessibleForFree": true, "license": "https://creativecommons.org/licenses/by/4.0/", "name": "Wavect verified publication claims", "url": "https://wavect.io/verified-claims.json" }, "url": "https://wavect.io/" }, { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "sameAs": [ "https://www.wikidata.org/wiki/Q139796365", "https://www.linkedin.com/in/wsdt", "https://github.com/wsdt" ], "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "sameAs": [ "https://www.wikidata.org/wiki/Q139796367", "https://www.linkedin.com/in/jocr77/", "https://github.com/jo-chris" ], "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/#website", "@type": "WebSite", "inLanguage": [ "en", "de", "es", "zh" ], "name": "Wavect", "potentialAction": { "@type": "SearchAction", "query-input": "required name=search_term_string", "target": { "@type": "EntryPoint", "urlTemplate": "https://wavect.io/search/?q={search_term_string}" } }, "publisher": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] }, "url": "https://wavect.io/" }, { "@id": "https://wavect.io/blog/phonely-alma-voice-llm-review/#webpage", "@type": "WebPage", "dateModified": "2026-09-06", "inLanguage": "en", "isPartOf": { "@id": "https://wavect.io/#website", "@type": "WebSite" }, "lastReviewed": "2026-09-06", "url": "https://wavect.io/blog/phonely-alma-voice-llm-review/" } ] } ``` ```json { "@context": "https://schema.org", "@type": "BlogPosting", "abstract": "Phonely's Alma is a specialist language model for structured phone conversations, not a general coding or long-horizon agent model. Phonely reports 182 ms time to first token, 206 ms P99, 379 ms full replies, a 0.867 phone-etiquette score and $0.55 per blended million tokens in a 200-transcript benchmark run on 21 August 2026. Those results make Alma a credible pilot candidate for scheduling, support, lead qualification and other repeatable call flows. They do not establish independent superiority or end-to-end caller latency because the vendor ran and graded the test, used sequential requests, and did not publish every prompt, transcript, judge, region or concurrency condition. Compare Alma with the incumbent on real calls, including accents, noise, interruptions, tool calls, policy adherence, escalation, P95 and P99 voice-to-voice latency, cost per successful call, data terms and fallback behavior. Adopt it only when the complete call outcome improves.", "articleBody": " Blog overview/AI and agents/Models and infrastructure Phonely Alma Review: Is the Voice LLM Ready for Production? TL;DR Phonely's Alma is a specialist language model for structured phone conversations, not a general coding or long-horizon agent model. Phonely reports 182 ms time to first token, 206 ms P99, 379 ms full replies, a 0.867 phone-etiquette score and $0.55 per blended million tokens in a 200-transcript benchmark run on 21 August 2026. Those results make Alma a credible pilot candidate for scheduling, support, lead qualification and other repeatable call flows. They do not establish independent superiority or end-to-end caller latency because the vendor ran and graded the test, used sequential requests, and did not publish every prompt, transcript, judge, region or concurrency condition. Compare Alma with the incumbent on real calls, including accents, noise, interruptions, tool calls, policy adherence, escalation, P95 and P99 voice-to-voice latency, cost per successful call, data terms and fallback behavior. Adopt it only when the complete call outcome improves. Alma is a credible specialist LLM for a narrow job: keeping structured phone conversations fast, concise and on task. Phonely reports 182 ms to the first token, 206 ms at P99, 379 ms for a full reply and $0.55 per blended million tokens. Those numbers earn Alma a production pilot. They do not prove that it is the best voice model for your callers. This review owns the model-selection decision inside a voice-agent stack. For the adjacent components, use our separate streaming speech-to-text production review and self-hosted versus API text-to-speech analysis. Keeping the three decisions separate prevents a fast LLM benchmark from being mistaken for a fast call. What is Phonely Alma? Alma is a language model optimized for repeatable phone workflows, conversational phrasing and tool calling. Phonely positions it for appointment booking, customer support, lead qualification, debt collection, lookups, transfers and other flows with a definable outcome. The company explicitly says it is not designed for coding, creative writing, open-ended multi-agent work or long-horizon planning. The term “voice-native LLM” needs care. Alma is priced in input and output tokens, and Phonely says it can work with existing transcription and speech-generation components. Evaluate it as the language-model stage unless your chosen deployment contract explicitly includes audio input, audio output, turn detection and telephony. A specialist text LLM can still be excellent for voice. It just does not remove the other clocks in the pipeline. What does Phonely claim for Alma? Vendor-reported Alma benchmark, measured 21 August 2026 MetricAlmaGPT-4.1GPT-5.6Buyer interpretation Time to first token182 ms490 ms997 msUseful LLM-stage responsiveness signal P99 response latency206 ms2.02 s2.84 sPromising tail result under the published test Full reply379 ms771 ms1.46 sSuggests short spoken answers are an optimization target Phone etiquette0.8670.7000.770Directionally useful, but rubric and judges matter Blended cost per 1M tokens$0.55$3.50Not pricedModel-token comparison, not total call cost The official Alma launch page says the comparison used 200 held-out call transcripts, identical prompts and grading, sequential requests from an AWS us-east-1 client, and five excluded warm-ups. Phonely also publishes shared pricing of $0.30 per million input tokens and $1.30 per million output tokens, priority pricing of $0.50 and $2.10, plus custom VPC or on-premises terms. Independence and trademarks Wavect publishes this page and is itself a provider, so we have a commercial interest in it. We are not affiliated with, endorsed by or partnered with the other companies named here, and all third-party company names, brands and trademarks are the property of their respective owners. Statements about other providers are taken from publicly available sources, primarily their own published pages, as of the review date shown on this page, and may have changed since. Please verify them directly before you decide. This page was written to the best of our knowledge and with the intent to remain objective. If you believe anything here is inaccurate or unfair, write to us and we will correct it: office@wavect.io What does the Alma benchmark prove? It proves that Alma deserves a head-to-head test for structured phone work. The vendor names the models, date, client region, sample size and several latency measures. Publishing a P99 is especially useful because one long silence can damage a call more than a strong median can rescue it. It does not establish general superiority. The evaluation is vendor-run and appears to use transcripts rather than complete live audio calls. The public page does not expose the full held-out set, prompt, phone-etiquette rubric, judge composition, confidence intervals, request lengths, cache state, rate tiers or concurrent-load profile. Sequential requests do not tell a", "articleSection": "Engineering", "author": { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "name": "Kevin Riedl", "sameAs": [ "https://www.wikidata.org/wiki/Q139796365", "https://www.linkedin.com/in/wsdt", "https://github.com/wsdt" ], "url": "https://wavect.io/team/kevin-riedl/" }, "citation": [ { "@type": "WebPage", "name": "official Alma launch page", "url": "https://www.phonely.ai/alma" }, { "@type": "WebPage", "name": "open PhoneBench evaluation design", "url": "https://www.pipecat.ai/benchmarks/phonebench-alpha-1" }, { "@type": "WebPage", "name": "τ-Voice study", "url": "https://arxiv.org/abs/2603.13686" } ], "dateModified": "2026-09-06", "datePublished": "2026-09-06", "description": "Phonely's Alma is a specialist language model for structured phone conversations, not a general coding or long-horizon agent model. Phonely reports 182 ms time to first token, 206 ms P99, 379 ms full replies, a 0.867 phone-etiquette score and $0.55 per blended million tokens in a 200-transcript benchmark run on 21 August 2026. Those results make Alma a credible pilot candidate for scheduling, support, lead qualification and other repeatable call flows. They do not establish independent superiority or end-to-end caller latency because the vendor ran and graded the test, used sequential requests, and did not publish every prompt, transcript, judge, region or concurrency condition. Compare Alma with the incumbent on real calls, including accents, noise, interruptions, tool calls, policy adherence, escalation, P95 and P99 voice-to-voice latency, cost per successful call, data terms and fallback behavior. Adopt it only when the complete call outcome improves.", "headline": "Phonely Alma Review: Is the Voice LLM Ready for Production?", "image": "https://wavect.io/img/blog/headers/header_phonely-alma-voice-llm-review.svg", "inLanguage": "en", "keywords": "Voice AI, LLM Evaluation", "mainEntityOfPage": { "@id": "https://wavect.io/blog/phonely-alma-voice-llm-review/", "@type": "WebPage" }, "publisher": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] }, "url": "https://wavect.io/blog/phonely-alma-voice-llm-review/", "wordCount": 2559 } ``` ```json { "@context": "https://schema.org", "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "item": "https://wavect.io/", "name": "Home", "position": 1 }, { "@type": "ListItem", "item": "https://wavect.io/blog/overview/", "name": "Blog overview", "position": 2 }, { "@type": "ListItem", "item": "https://wavect.io/blog/topics/ai-agents/", "name": "AI and agents", "position": 3 }, { "@type": "ListItem", "item": "https://wavect.io/blog/clusters/models-infrastructure/", "name": "Models and infrastructure", "position": 4 }, { "@type": "ListItem", "item": "https://wavect.io/blog/phonely-alma-voice-llm-review/", "name": "Phonely Alma Review: Voice LLM Cost & Latency", "position": 5 } ] } ``` ```json { "@context": "https://schema.org", "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Alma is a specialist language model built for structured phone conversations. Phonely positions it for scheduling, support, qualification, collections, tool calls and concise conversational behavior, not coding or long-horizon general agent work." }, "name": "What is Phonely Alma?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Do not assume that from the voice-native label. Phonely prices Alma in text tokens and says it works with existing transcription and text-to-speech components. Evaluate the contracted product boundary and measure the whole audio pipeline." }, "name": "Is Alma a speech-to-speech model?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "In Phonely’s vendor-run test of 200 held-out call transcripts, Alma reported 182 ms time to first token versus 490 ms for GPT-4.1, plus 206 ms versus 2.02 seconds at P99. Reproduce the result under your prompts, region, tier, concurrency and full call stack." }, "name": "Is Alma really faster than GPT-4.1?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "On 6 September 2026, Phonely listed shared rates of $0.30 per million input tokens and $1.30 per million output tokens, and priority rates of $0.50 and $2.10. It reported $0.55 per blended million tokens in its benchmark. VPC and on-premises pricing is custom." }, "name": "How much does Alma cost?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Measure correct call outcomes, tool-state accuracy, policy adherence, human-rated conversation quality, interruption behavior, voice-to-voice P50, P95 and P99, failures, fallback success and total cost per accepted resolution." }, "name": "What should an Alma pilot measure?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "No. Alma fits repeatable, definable phone flows. Keep or compare a general model for open-ended reasoning and long tasks. Switch only after a controlled test shows better complete-call outcomes without new safety, reliability or data risks." }, "name": "Should every voice agent switch to Alma?" } ] } ```