# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/story/49823348 # Mercury 2\.5 LLM hits 770 tokens per second 118 points · 68 comments [Full discussion]() [Read original]() Category: [Infrastructure & Efficiency]() ## Skept-o-meter & Hotness Skept\-o\-meter: High\. Estimated from 10 comments\. 15 comments for the summary\. Peak rank: \#5 Time in Top 10: 24\.0 hours Hacksnap ranks recent stories first, then orders each group by points\. Peak rank uses all retained observations\. Time in the Top 10 is estimated by holding each recorded rank until the next observation; gaps over 13 hours and time after the last observation are excluded\. Movement between observations is unknown\. 224 recorded rank observations from 2026\-09\-24T12:00:45\.621589\+00:00 to 2026\-10\-10T23:01:00\.95027\+00:00\. Hotness — latest 168 recorded Hacksnap ranks: 2026\-09\-30T18:03:34\.83179\+00:00: rank \#131 2026\-09\-30T19:02:02\.291882\+00:00: rank \#131 2026\-09\-30T20:02:49\.392147\+00:00: rank \#132 2026\-09\-30T21:03:42\.990832\+00:00: rank \#131 2026\-09\-30T22:02:11\.294387\+00:00: rank \#132 2026\-09\-30T23:01:52\.722953\+00:00: rank \#132 2026\-10\-01T08:02:24\.797064\+00:00: rank \#132 2026\-10\-01T09:01:35\.899477\+00:00: rank \#135 2026\-10\-01T10:01:12\.485692\+00:00: rank \#135 2026\-10\-01T11:00:47\.508128\+00:00: rank \#135 2026\-10\-01T12:00:42\.477917\+00:00: rank \#134 2026\-10\-01T13:00:51\.124852\+00:00: rank \#134 2026\-10\-01T14:00:59\.01153\+00:00: rank \#135 2026\-10\-01T15:00:51\.048047\+00:00: rank \#136 2026\-10\-01T16:00:53\.034654\+00:00: rank \#137 2026\-10\-01T16:36:37\.180071\+00:00: rank \#138 2026\-10\-01T17:00:23\.382263\+00:00: rank \#138 2026\-10\-01T18:01:44\.046011\+00:00: rank \#139 2026\-10\-01T19:01:47\.420738\+00:00: rank \#140 2026\-10\-01T20:01:39\.585083\+00:00: rank \#140 2026\-10\-01T21:01:45\.872085\+00:00: rank \#141 2026\-10\-01T22:01:37\.214899\+00:00: rank \#140 2026\-10\-01T23:04:11\.395288\+00:00: rank \#143 2026\-10\-02T08:02:26\.819307\+00:00: rank \#144 2026\-10\-02T09:02:51\.51649\+00:00: rank \#146 2026\-10\-02T10:02:12\.346099\+00:00: rank \#147 2026\-10\-02T11:02:22\.794376\+00:00: rank \#147 2026\-10\-02T12:01:14\.681578\+00:00: rank \#147 2026\-10\-02T13:02:29\.511591\+00:00: rank \#147 2026\-10\-02T14:01:40\.936738\+00:00: rank \#147 2026\-10\-02T15:02:30\.225476\+00:00: rank \#148 2026\-10\-02T16:02:47\.060695\+00:00: rank \#148 2026\-10\-02T17:02:31\.351493\+00:00: rank \#150 2026\-10\-02T18:01:22\.271899\+00:00: rank \#151 2026\-10\-02T19:01:23\.681417\+00:00: rank \#152 2026\-10\-02T20:01:33\.801055\+00:00: rank \#154 2026\-10\-02T21:02:55\.453316\+00:00: rank \#157 2026\-10\-02T22:01:50\.233139\+00:00: rank \#158 2026\-10\-02T23:01:20\.268516\+00:00: rank \#157 2026\-10\-03T08:01:30\.186188\+00:00: rank \#159 2026\-10\-03T09:01:57\.91467\+00:00: rank \#159 2026\-10\-03T10:00:55\.732001\+00:00: rank \#159 2026\-10\-03T11:01:22\.538596\+00:00: rank \#159 2026\-10\-03T12:01:08\.789384\+00:00: rank \#160 2026\-10\-03T13:00:35\.273559\+00:00: rank \#160 2026\-10\-03T14:00:48\.425606\+00:00: rank \#160 2026\-10\-03T15:01:04\.472963\+00:00: rank \#160 2026\-10\-03T16:01:56\.56175\+00:00: rank \#161 2026\-10\-03T17:01:36\.651036\+00:00: rank \#159 2026\-10\-03T18:01:58\.24207\+00:00: rank \#159 2026\-10\-03T19:01:19\.196589\+00:00: rank \#159 2026\-10\-03T20:00:57\.313759\+00:00: rank \#159 2026\-10\-03T21:01:04\.44439\+00:00: rank \#159 2026\-10\-03T21:20:01\.585706\+00:00: rank \#159 2026\-10\-03T22:00:15\.503126\+00:00: rank \#159 2026\-10\-03T23:00:55\.584832\+00:00: rank \#159 2026\-10\-04T08:01:00\.327459\+00:00: rank \#159 2026\-10\-04T09:00:59\.321023\+00:00: rank \#160 2026\-10\-04T10:00:42\.975336\+00:00: rank \#160 2026\-10\-04T11:01:03\.76577\+00:00: rank \#160 2026\-10\-04T12:01:45\.307985\+00:00: rank \#161 2026\-10\-04T13:00:49\.176267\+00:00: rank \#161 2026\-10\-04T14:00:56\.952543\+00:00: rank \#161 2026\-10\-04T15:00:54\.739637\+00:00: rank \#162 2026\-10\-04T16:00:51\.403054\+00:00: rank \#163 2026\-10\-04T17:00:43\.166713\+00:00: rank \#163 2026\-10\-04T18:00:46\.315535\+00:00: rank \#162 2026\-10\-04T19:00:54\.768841\+00:00: rank \#162 2026\-10\-04T20:00:51\.070442\+00:00: rank \#163 2026\-10\-04T21:00:28\.550532\+00:00: rank \#162 2026\-10\-04T22:01:36\.217651\+00:00: rank \#164 2026\-10\-04T23:01:09\.400767\+00:00: rank \#165 2026\-10\-05T08:01:51\.387003\+00:00: rank \#165 2026\-10\-05T09:01:40\.528475\+00:00: rank \#165 2026\-10\-05T10:00:56\.290772\+00:00: rank \#165 2026\-10\-05T11:01:05\.919518\+00:00: rank \#165 2026\-10\-05T12:00:56\.590881\+00:00: rank \#165 2026\-10\-05T13:01:05\.197156\+00:00: rank \#167 2026\-10\-05T14:01:04\.085718\+00:00: rank \#167 2026\-10\-05T15:00:50\.814571\+00:00: rank \#167 2026\-10\-05T16:02:09\.54139\+00:00: rank \#167 2026\-10\-05T17:00:35\.336076\+00:00: rank \#167 2026\-10\-05T18:00:20\.445491\+00:00: rank \#167 2026\-10\-05T19:02:43\.604975\+00:00: rank \#168 2026\-10\-05T20:01:09\.777459\+00:00: rank \#168 2026\-10\-05T21:01:47\.650671\+00:00: rank \#169 2026\-10\-05T22:02:59\.93099\+00:00: rank \#170 2026\-10\-06T08:02:11\.210914\+00:00: rank \#170 2026\-10\-06T08:02:14\.354697\+00:00: rank \#170 2026\-10\-06T09:01:21\.688765\+00:00: rank \#172 2026\-10\-06T10:00:58\.047773\+00:00: rank \#172 2026\-10\-06T11:00:41\.689929\+00:00: rank \#173 2026\-10\-06T12:00:41\.7705\+00:00: rank \#174 2026\-10\-06T13:01:24\.509519\+00:00: rank \#174 2026\-10\-06T14:00:26\.950817\+00:00: rank \#174 2026\-10\-06T15:00:49\.066414\+00:00: rank \#176 2026\-10\-06T16:00:58\.166747\+00:00: rank \#178 2026\-10\-06T17:01:17\.009511\+00:00: rank \#178 2026\-10\-06T18:01:43\.791429\+00:00: rank \#179 2026\-10\-06T19:00:46\.048609\+00:00: rank \#179 2026\-10\-06T20:01:14\.687294\+00:00: rank \#180 2026\-10\-06T21:02:13\.414382\+00:00: rank \#180 2026\-10\-06T22:00:40\.926271\+00:00: rank \#179 2026\-10\-06T23:03:26\.867445\+00:00: rank \#181 2026\-10\-07T08:01:41\.339804\+00:00: rank \#182 2026\-10\-07T09:02:38\.026986\+00:00: rank \#184 2026\-10\-07T10:01:35\.069531\+00:00: rank \#184 2026\-10\-07T11:02:07\.770084\+00:00: rank \#184 2026\-10\-07T12:01:14\.996697\+00:00: rank \#183 2026\-10\-07T13:00:47\.45912\+00:00: rank \#182 2026\-10\-07T14:01:18\.307739\+00:00: rank \#182 2026\-10\-07T15:01:57\.541362\+00:00: rank \#185 2026\-10\-07T16:01:28\.474022\+00:00: rank \#185 2026\-10\-07T17:01:15\.036908\+00:00: rank \#186 2026\-10\-07T18:00:53\.300079\+00:00: rank \#186 2026\-10\-07T19:01:49\.456657\+00:00: rank \#187 2026\-10\-07T20:01:59\.032487\+00:00: rank \#187 2026\-10\-07T21:05:36\.655392\+00:00: rank \#189 2026\-10\-07T22:02:27\.105241\+00:00: rank \#190 2026\-10\-07T23:01:17\.452894\+00:00: rank \#190 2026\-10\-08T08:03:55\.083808\+00:00: rank \#190 2026\-10\-08T09:03:58\.68898\+00:00: rank \#191 2026\-10\-08T10:04:05\.364235\+00:00: rank \#192 2026\-10\-08T11:02:04\.410307\+00:00: rank \#191 2026\-10\-08T12:02:36\.39618\+00:00: rank \#191 2026\-10\-08T13:03:09\.206401\+00:00: rank \#191 2026\-10\-08T14:03:30\.113384\+00:00: rank \#192 2026\-10\-08T15:02:54\.444758\+00:00: rank \#190 2026\-10\-08T16:03:13\.067807\+00:00: rank \#189 2026\-10\-08T17:03:47\.92465\+00:00: rank \#188 2026\-10\-08T18:01:50\.360431\+00:00: rank \#188 2026\-10\-08T19:02:55\.719274\+00:00: rank \#189 2026\-10\-08T20:01:36\.359022\+00:00: rank \#189 2026\-10\-08T21:03:34\.000719\+00:00: rank \#191 2026\-10\-08T22:01:13\.459732\+00:00: rank \#191 2026\-10\-08T23:01:25\.318392\+00:00: rank \#190 2026\-10\-09T08:01:53\.314021\+00:00: rank \#191 2026\-10\-09T09:02:59\.93426\+00:00: rank \#192 2026\-10\-09T10:02:22\.652527\+00:00: rank \#192 2026\-10\-09T11:02:04\.807282\+00:00: rank \#192 2026\-10\-09T12:02:16\.769295\+00:00: rank \#192 2026\-10\-09T13:02:30\.35034\+00:00: rank \#194 2026\-10\-09T14:01:34\.215134\+00:00: rank \#194 2026\-10\-09T15:01:27\.564392\+00:00: rank \#194 2026\-10\-09T16:02:06\.561056\+00:00: rank \#195 2026\-10\-09T17:03:06\.673648\+00:00: rank \#197 2026\-10\-09T18:01:02\.851052\+00:00: rank \#197 2026\-10\-09T19:01:11\.250291\+00:00: rank \#198 2026\-10\-09T20:01:32\.91907\+00:00: rank \#198 2026\-10\-09T21:02:08\.376144\+00:00: rank \#200 2026\-10\-09T22:01:19\.253075\+00:00: rank \#202 2026\-10\-09T23:01:31\.990912\+00:00: rank \#202 2026\-10\-10T08:01:00\.593376\+00:00: rank \#201 2026\-10\-10T09:01:45\.527975\+00:00: rank \#204 2026\-10\-10T10:00:53\.799353\+00:00: rank \#204 2026\-10\-10T11:00:35\.65851\+00:00: rank \#203 2026\-10\-10T12:01:11\.169647\+00:00: rank \#203 2026\-10\-10T13:00:54\.008377\+00:00: rank \#203 2026\-10\-10T14:01:43\.208766\+00:00: rank \#205 2026\-10\-10T15:01:24\.825188\+00:00: rank \#205 2026\-10\-10T16:01:10\.213328\+00:00: rank \#204 2026\-10\-10T17:00:48\.84189\+00:00: rank \#203 2026\-10\-10T18:00:57\.55132\+00:00: rank \#203 2026\-10\-10T19:00:52\.379463\+00:00: rank \#202 2026\-10\-10T20:00:21\.993912\+00:00: rank \#202 2026\-10\-10T21:01:10\.657738\+00:00: rank \#204 2026\-10\-10T22:00:50\.433569\+00:00: rank \#203 2026\-10\-10T23:01:00\.95027\+00:00: rank \#204 Mercury 2\.5's headline speed is real, but the sampled discussion largely doubts it matters given below\-average intelligence, cheaper or faster alternatives, and poor hands\-on reports; the main defense is that raw speed itself is the product\. ## The brief Artificial Analysis lists Mercury 2\.5, a proprietary reasoning model from Inception released September 8, 2026, as extremely fast but below average in intelligence\. It reports 780\.8 output tokens per second, a score of 12 on the Artificial Analysis Intelligence Index, $0\.25 per 1M input tokens and $0\.75 per 1M output tokens, a 260k\-token context window, text\-only input/output, and a 2\.91s time to first token\. - Mercury 2\.5 is a proprietary reasoning model from Inception, released September 8, 2026, with text\-only input and output and a 260k\-token context window\. - It generates 780\.8 output tokens per second, ranked \#2 of 173 models and well above the reported median of 110\.3 t/s for comparable reasoning models\. - It scores 12 on the Artificial Analysis Intelligence Index, ranked \#89 of 173 and described as below average among comparable models\. - Pricing is $0\.25 per 1M input tokens and $0\.75 per 1M output tokens, with a 90% cache discount and an estimated $0\.06 cost per Intelligence Index task\. - It is fairly concise on the Intelligence Index, generating 35M output tokens versus a median of 85M\. - Its time to first token is 2\.91s, somewhat higher than the median 2\.22s, and it does not support image input or multimodal use\. ## Discussion themes ### Speed does not fix weak results Several commenters argue that fast output is not useful if the model is far behind frontier models or produces bad answers\. One says speed means nothing when the model finishes almost dead last against frontier AI companies, another says bad results make speed irrelevant, and another summarizes it as 'stupid and fast\.' Sources: [Comment 49823424]() · [Comment 49825273]() · [Comment 49825006]() ### Pricing and value are questioned A commenter says $0\.25/$0\.75 pricing is already above reputable inference providers for open\-weight models like DeepSeek v4 Flash or Qwen 3\.8\-Flash\-Next, and suspects Mercury 2\.5 is less capable than Laguna S 2\.1\. Another says the speed is less impressive now because other providers offer DeepSeek Flash around 600 t/s and K3/GLM in the hundreds\. Sources: [Comment 49823863]() · [Comment 49825901]() ### Faster alternatives exist, but with caveats One commenter points to Cerebras gpt\-oss\-120b at 1400 t/s as 'just as smart' in ranking\. A reply warns that using gpt\-oss\-120b over Cerebras is broken for tool calls, fails to end thinking blocks, and has no cache pricing, making it very expensive\. Another notes Kimi 2\.6 at 1000 t/s but with a long waitlist and high minimum spend\. Sources: [Comment 49824041]() · [Comment 49825098]() · [Comment 49824707]() ### Diffusion LLM approach seen as a dead end A commenter argues the diffusion LLM approach is a dead end, noting that frontier labs like Google experimented with it but did not invest further even for speed\- and cost\-sensitive small models, and that it remains unclear which use cases put it on the Pareto frontier\. Sources: [Comment 49824229]() ### Hands\-on reports are poor One commenter says they tried Mercury 2\.5 for many tasks but found it not there, roughly on par with a 14B model at best, and worse than GPT\-OSS\-20B even for basic tasks\. Another reports an absurdly bad constrained French\-generation result and wonders about tradeoffs with larger or smarter models\. A third says such speeds are mainly useful for customer service or transactional workflows, not coding, because fast code generation still leaves hours of review\. Sources: [Comment 49825905]() · [Comment 49828064]() · [Comment 49826304]() ### Counterpoint: raw speed is the product Some commenters push back that speed itself is the point\. One says simply 'The point is the speed,' another says 'it's stupid fast,' and another wishes for a new Taalas\-style chip because smaller models are now good enough to make bursts of tokens useful\. Sources: [Comment 49824890]() · [Comment 49824690]() · [Comment 49826041]() ## Sources & coverage AI-generated summary · 2026\-09\-24T12:00:43\.540598\+00:00 Based on 15 of 15 usable stored comments, selected by depth and branch activity. This is a sample of the discussion. Article text may also be shortened. Generated using deepseek\-ai/DeepSeek\-V4\.1\-Flash. Check the linked sources for full context.