[HN Gopher] Speed up responses with fast mode
       ___________________________________________________________________
        
       Speed up responses with fast mode
        
       Author : surprisetalk
       Score  : 73 points
       Date   : 2026-02-07 18:08 UTC (4 hours ago)
        
 (HTM) web link (code.claude.com)
 (TXT) w3m dump (code.claude.com)
        
       | thehamkercat wrote:
       | Interesting, output price is insane/Mtok
        
       | 1123581321 wrote:
       | Could be a use for the $50 extra usage credit. It requires extra
       | usage to be enabled.
       | 
       | > Fast mode usage is billed directly to extra usage, even if you
       | have remaining usage on your plan. This means fast mode tokens do
       | not count against your plan's included usage and are charged at
       | the fast mode rate from the first token.
        
         | minimaxir wrote:
         | After exceeding the increasingly shrinking session limit with
         | Opus 4.6, I continued with the extra usage only for a few
         | minutes and it consumed about $10 of the credit.
         | 
         | I can't imagine how quickly this Fast Mode goes through credit.
        
         | arcanemachiner wrote:
         | It has to be. The timing is just too close.
        
       | simonw wrote:
       | The one question I have that isn't answered by the page is _how
       | much faster_?
       | 
       | Obviously they can't make promises but I'd still like a rough
       | indication of how much this might improve the speed of responses.
        
         | l1n wrote:
         | 2.5x faster or so
         | (https://x.com/claudeai/status/2020207322124132504).
        
           | zurfer wrote:
           | 6x more expensive
        
         | scosman wrote:
         | Yeah is this cerebras/groq speed, or I just skip the queue?
        
       | krm01 wrote:
       | Will this mean that when cost is more important than latency that
       | replies will now take longer?
       | 
       | I'm not in favor of the ad model chatgpt proposes. But business
       | models like these suffer from similar traps.
       | 
       | If it works for them, then the logical next step is to convert
       | more to use fast mode. Which naturally means to slow things down
       | for those that didn't pick/pay for fast mode.
       | 
       | We've seen it with iPhones being slowed down to make the newer
       | model seem faster.
       | 
       | Not saying it'll happen. I love Claude. But these business models
       | almost always invite dark patterns in order to move the bottom
       | line.
        
       | speedping wrote:
       | > $30/150 MTok Umm no thank you
        
       | pedropaulovc wrote:
       | Where is this perf gain coming from? Running on TPUs?
        
         | AnotherGoodName wrote:
         | AI data centers are a whole lot of pipelines pumping data
         | around utilizing queues. They want those expensive power hungry
         | cards near 100% utilized at all times. So they have a queue of
         | jobs on each system ready to run, feeding into the GPU memory
         | as fast as completed jobs are read out of memory (and passed
         | into the next stage) and they aim to have enough backlog in
         | these queues to keep the pipeline full. You see responses in
         | seconds but at the data center you're request was broken into
         | jobs, passed around into queues, processed in an orderly manner
         | and pieced back together.
         | 
         | With fast mode you're literally skipping the queue. An outcome
         | of all of this is that for the rest of us the responses will
         | become slower the more people use this 'fast' option.
         | 
         | I do suspect they'll also soon have a slow option for those
         | that have Claude doing things overnight with no real care for
         | latency of the responses. The ultimate goal is pipelines of
         | data hitting 100% hardware utilization at all times.
        
       | pronik wrote:
       | While it's an excellent way to make more money in the moment, I
       | think this might become a standard no-extra-cost feature in
       | several months (see Opus becoming way cheaper and a default model
       | within months). Mental load management while using agents will
       | become even more important it seems.
        
         | giancarlostoro wrote:
         | Yeah especially once they make an even faster fast mode.
        
         | falloutx wrote:
         | Why would they cut a money making feature? In fact I am already
         | imagining them asking for speed ransom every time you are in a
         | pinch, some extra context space will also become buyable.
         | Anthropic is in a penny pincher phase right now and they will
         | try to milk everything. Watch them add micro transactions too.
        
       | Nition wrote:
       | Note that you can't use this mode to get the most out of a
       | subscription - they say it's always charged as extra usage:
       | 
       | > Fast mode usage is billed directly to extra usage, even if you
       | have remaining usage on your plan. This means fast mode tokens do
       | not count against your plan's included usage and are charged at
       | the fast mode rate from the first token.
       | 
       | Although if you visit the Usage screen right now, there's a deal
       | you can claim for $50 free extra usage this month.
        
       | IMTDb wrote:
       | I'm curious what's behind the speed improvements. It seems
       | unlikely it's _just_ prioritization, so what else is changing? Is
       | it new hardware (a la Groq or Cerebras)? That seems plausible,
       | especially since it isn't available on some cloud providers.
       | 
       | Also wondering whether we'll soon see separate "speed" vs
       | "cleverness" pricing on other LLM providers too.
        
         | pshirshov wrote:
         | > so what else is changing?
         | 
         | Let me guess. Quantization?
        
         | sothatsit wrote:
         | There are a lot of knobs they could tweak. Newer hardware and
         | traffic prioritisation would both make a lot of sense. But they
         | could also lower batching windows to decrease queueing time at
         | the cost of lower throughput, or keep the KV cache in GPU
         | memory at the expense of reducing the number of users they can
         | serve from each GPU node.
        
         | Nition wrote:
         | I wonder if they might have mostly implemented this for
         | themselves to use internally, and it is just prioritization but
         | they don't expect too many others to pay the high cost.
        
           | sothatsit wrote:
           | Roon said as much here [0]:
           | 
           | > codex-5.2 is really amazing but using it from my personal
           | and not work account over the weekend taught me some user
           | empathy lol it's a bit slow
           | 
           | [0] https://nitter.net/tszzl/status/2016338961040548123
        
         | jstummbillig wrote:
         | > It seems unlikely it's just prioritization
         | 
         | Why does this seem unlikely? I have no doubt they are
         | optimizing all the time, including inference speed, but why
         | could this particular lever not entirely be driven by skipping
         | the queue? It's an easy way to generate more money.
        
           | singpolyma3 wrote:
           | Until everyone buys it. Like fast pass at an amusement park
           | where the fast line is still two hours long
        
             | servercobra wrote:
             | It's a good way to squeeze extra out of a bunch of people
             | without actually raising prices.
        
             | sothatsit wrote:
             | At 6x the cost, and it requiring you to pay full API
             | pricing, I don't think this is going to be a concern.
        
           | AnotherGoodName wrote:
           | Yes it's 100% prioritization. Through that it's also likely
           | running on more GPUs at once but that's an artifact of
           | prioritization at the datacenter level. Any task coming into
           | an AI datacenter atm is split into fairly fined grained
           | chunks of work and added to queues to be processed.
           | 
           | When you add a job with high priority all those chunks will
           | be processed off the queue first by each and every GPU that
           | frees up. It probably leads to more parallelism but... it's
           | the prioritization that led to this happening. It's better to
           | think of this as prioritization of your job leading to the
           | perf improvement.
           | 
           | Here's a good blog for anyone interested which talks about
           | prioritization and job scheduling. It's not quite at the
           | datacenter level but the concepts are the same. Basically
           | everything is thought of as a pipeline. All training jobs are
           | low pri (they take months to complete in any case), customer
           | requests are mid pri and then there's options for high pri.
           | Everything in an AI datacenter is thought of in terms of
           | 'flow'. Are there any bottlenecks? Are the pipelines always
           | full and the expensive hardware always 100% utilized? Are the
           | queues backlogs big enough to ensure full utilization at
           | every stage?
           | 
           | https://www.aleksagordic.com/blog/vllm
        
         | kingstnap wrote:
         | It comes from batching and multiple streams on a GPU. More
         | people sharing 1 GPU makes everyone run slower but increases
         | overall token throughput.
         | 
         | Mathematically it comes from the fact that this transformer
         | block is this parallel algorithm. If you batch harder, increase
         | parallelism, you can get higher tokens/s. But you get less
         | throughput. Simultaneously there is also this dial that you can
         | speculatively decode harder with fewer users.
         | 
         | Its true for basically all hardware and most models. You can
         | draw this Pareto curve of how much throughput per GPU vs how
         | many tokens per second per stream. More tokens/s less total
         | throughput.
         | 
         | See this graph for actual numbers:
         | 
         | Token Throughput per GPU vs. Interactivity gpt-oss 120B * FP4 *
         | 1K / 8K * Source: SemiAnalysis InferenceMAX(tm)
         | 
         | https://inferencemax.semianalysis.com/
        
         | re-thc wrote:
         | Nvidia GB300 i.e. Blackwell.
        
       | solidasparagus wrote:
       | I pay $200 a month and don't get any included access to this?
       | Ridiculous
        
         | pedropaulovc wrote:
         | Well, you can burn your $50 bonus on it
        
         | bakugo wrote:
         | The API price is 6x that of normal Opus, so look forward to a
         | new $1200/mo subscription that gives you the same amount of
         | usage if you need the extra speed.
        
           | MuffinFlavored wrote:
           | I always wondered this, is this true/does the math come out
           | to be really that bad? 6x?
           | 
           | Is the writing on the wall for $100-$200/mo users that, it's
           | basically known-subsidized for now and $400/mo+ is coming
           | sooner than we think?
           | 
           | Are they getting us all hooked and then going to raise it in
           | the future, or will inference prices go down to offset?
        
         | kingforaday wrote:
         | ..But it says "Available to all Claude Code users on
         | subscription plans (Pro/Max/Team/Enterprise) and Claude
         | Console."
         | 
         | Is this wrong?
        
           | behindsight wrote:
           | It's explicitly called out as excluded in the blue info
           | bubble they have there.
           | 
           | > Fast mode usage is billed directly to extra usage, even if
           | you have remaining usage on your plan. This means fast mode
           | tokens do not count against your plan's included usage and
           | are charged at the fast mode rate from the first token.
           | 
           | https://code.claude.com/docs/en/fast-mode#requirements
        
           | sothatsit wrote:
           | I think this is just worded in a misleading way. It's
           | available to all users, but it's not included as part of the
           | plan.
        
       | hmokiguess wrote:
       | Give me a slow mode that's cheaper instead lol
        
       | jhack wrote:
       | The pricing on this is absolutely nuts.
        
         | nick49488171 wrote:
         | For us mere mortals, how fast does a normal developer for
         | through a MTok. How about a good power user?
        
           | snowfield wrote:
           | A developer can blast millions of tokens in minutes. When you
           | have a context size of 250k that's just 4 queries. But with
           | tool usage and subsequent calls etc it can easily just do
           | many millions in one request
           | 
           | But if you just ask a question or something it'll take a
           | while to spend a million tokens...
        
             | nick49488171 wrote:
             | Seems like an opportunity to condense the context into
             | 'documentation' level and only load the full text/code for
             | files that expect to be edited?
        
       | clbrmbr wrote:
       | I'd love to hear from engineers who find that faster speed is a
       | big unlock for them.
       | 
       | The deadline piece is really interesting. I suppose there's a lot
       | of people now who are basically limited by how fast their agents
       | can run and on very aggressive timelines with funders breathing
       | down their necks?
        
         | sothatsit wrote:
         | If it could help avoid you needing to context switch between
         | multiple agents, that could be a big mental load win.
        
         | throw310822 wrote:
         | The idea of development teams bottlenecked by agent speed
         | rather than people, ideas, strategy, etc. gives me some strange
         | vibes.
        
         | Aurornis wrote:
         | > I'd love to hear from engineers who find that faster speed is
         | a big unlock for them.
         | 
         | How would it not be a big unlock? If the answers were instant I
         | could stay focused and iterate even faster instead of having a
         | back-and-forth.
         | 
         | Right now even medium requests can take 1-2 minutes and
         | significant work can take even longer. I can usually make some
         | progress on a code review, read more docs, or do a tiny chunk
         | of productive work but the constant context switching back and
         | forth every 60s is draining.
        
       | maz1b wrote:
       | AFAIK, they don't have any deals or partnerships with Groq or
       | Cerebras or any of those kinds of companies.. so how did they do
       | this?
        
         | hendersoon wrote:
         | Could well be running on Google TPUs.
        
         | tcdent wrote:
         | Inference is run on shared hardware already, so they're not
         | giving you the full bandwidth of the system by default. This
         | most likely just allocates more resources to your request.
        
       | esafak wrote:
       | It's a good way to address the price insensitive segment. As long
       | as they don't slow down the rest, good move.
        
       | paxys wrote:
       | Looking at the "Decide when to use fast mode", it seems the
       | future they want is:
       | 
       | - Long running autonomous agents and background tasks use regular
       | processing.
       | 
       | - "Human in the loop" scenarios use fast mode.
       | 
       | Which makes perfect sense, but the question is - does the billing
       | also make sense?
        
       | l5870uoo9y wrote:
       | It doesn't say how much faster it is but from my experience with
       | OpenAI's "service_tier=priority" option on SQLAI.ai is that it's
       | twice as fast.
        
       | simianwords wrote:
       | Whatever optimisation is going on is at the hardware level since
       | the fast option persists in a session.
        
       | aabhay wrote:
       | What is "$30/150MTok"? Claude Opus 4.6 is normally priced at
       | "$25/MTok". Am I just reading it wrong or is this a typo?
       | 
       | EDIT: I understand now. $30 for input, $150 for output. Very
       | confusing wording. That's insanely expensive!
        
         | stavros wrote:
         | Yeah I don't understand. Is it actually saying that fast mode
         | is ten times more expensive than normal mode? I cannot be
         | reading this right.
        
       | niobe wrote:
       | So fast mode uses more tokens, in direct opposition to Gemini
       | where fast 'mode' means less. One more piece of useless knowledge
       | to remember.
        
         | Sol- wrote:
         | I don't think this is the case, according to the docs, right?
         | The effort level will use fewer tokens, but the independent
         | fast mode just somehow seems to use some higher priority
         | infrastructure to serve your requests.
        
         | Aurornis wrote:
         | You're comparing two different things. It's not useless
         | knowledge, it's something you need to understand.
         | 
         | Opus fast mode is routed to different servers with different
         | tuning that prioritizes individual response throughput. Same
         | model served differently. Same response, just delivered faster.
         | 
         | The Gemini fast mode is a different model (most likely) with
         | different levels of thinking applied. Very different response.
        
       | rustyhancock wrote:
       | At this point why don't we just CNAME HN to the Claude marketing
       | blog?
        
       | AnotherGoodName wrote:
       | But waiting for the agent to finish is my 2026 equivalent of
       | "compiling!"
       | 
       | https://xkcd.com/303/
        
       | thisisauserid wrote:
       | Instead of better/cheaper/faster you just the the last one?
       | 
       | Back to Gemini.
        
       | jawon wrote:
       | I was thinking about inhouse model inference speeds at frontier
       | labs like Anthropic and OpenAI after reading the "Claude built a
       | C compiler" article.
       | 
       | Having higher inference speed would be an advantage, especially
       | if you're trying to eat all the software and services.
       | 
       | Anthropic offering 2.5x makes me assume they have 5x or 10x
       | themselves.
       | 
       | In the predicted nightmare future where everything happens via
       | agents negotiating with agents, the side with the most compute,
       | and the fastest compute, is going to steamroll everyone.
        
         | crowbahr wrote:
         | Where on earth are you getting these numbers? Why would a SaaS
         | company that is fighting for market dominance withhold 10x
         | performance if they had it? Where are you getting 2.5x?
         | 
         | This is such bizarre magical thinking, borderline
         | conspiratorial.
         | 
         | There is no reason to believe any of the big AI players are
         | serving anything less than the best trade off of stability and
         | speed that they can possibly muster, especially when their cost
         | ratios are so bad.
        
           | jawon wrote:
           | Not magical thinking, not conspiratorial, just hypothetical.
           | 
           | Just because you can't afford to 10x all your customers'
           | inference doesn't mean you can't afford to 10x your inhouse
           | inference.
           | 
           | And 2.5x is from Anthropic's latest offering. But it costs
           | you 6x normal API pricing.
        
             | jawon wrote:
             | Also, from a comment in another thread, from roon, who
             | works at OpenAI:
             | 
             | > codex-5.2 is really amazing but using it from my personal
             | and not work account over the weekend taught me some user
             | empathy lol it's a bit slow
             | 
             | [0] https://nitter.net/tszzl/status/2016338961040548123
        
         | falloutx wrote:
         | Thats also called slowing down default experience so users have
         | to pay more for the fast mode. I think its the first time we
         | are seeing blatant speed ransoms in the LLMs.
        
           | throw310822 wrote:
           | Slowing down respect to what?
        
             | falloutx wrote:
             | Slowing down with respect to original speed of response.
             | Basically what we used to get few months back and what is
             | the best possible experience.
        
               | throw310822 wrote:
               | There is no "original speed of response". The more
               | resources you pour in, the faster it goes.
        
               | falloutx wrote:
               | Watch them decrease resources for the normal mode so
               | people are penny pinched into using fast mode.
        
               | throw310822 wrote:
               | Seriously, thinking at the price structure of this (6x
               | the price for 2.5x the speed, if that's correct) it seems
               | to target something like real time applications with very
               | small context. Maybe vocal assistants? I guess that if
               | you're doing development it makes more sense to
               | parallelize over more agents rather than paying that much
               | for a modest increase in speed.
        
           | Aurornis wrote:
           | That's not how this works. LLM serving at scale processes
           | multiple requests in parallel for efficiency. Reduce the
           | parallelism and you can process individual requests faster,
           | but the overall number of tokens processed is lower.
        
             | falloutx wrote:
             | They can now easily decrease the speed for the normal mode,
             | and then users will have to pay more for fast mode.
        
               | Aurornis wrote:
               | Do you have any evidence that this is happening? Or is it
               | just a hypothetical threat you're proposing?
               | 
               | These companies aren't operating in a vacuum. Most of
               | their users could change providers quickly if they
               | started degrading their service.
        
               | falloutx wrote:
               | They have contracts with companies, and those companies
               | wont be able to change quickly. By the time those
               | contracts will come back for renewals it will already be
               | too late, their code becoming completely unreadable by
               | humans. Individual devs can move quickly but companies
               | don't.
        
               | kolinko wrote:
               | Are you at all familiar with the architecture of systems
               | like theirs?
               | 
               | The reason people don't jump to your conclusion here (and
               | why you get downvoted) is that for anyone familiar with
               | how this is orchestrated on the backend it's obvious that
               | they don't need to do artificial slowdowns.
        
               | falloutx wrote:
               | I am familiar with the business model. This is clear
               | indication of what their future plan is.
               | 
               | Also, I just pointed out at the business issue, just
               | raising a point which was not raised here. Just want
               | people to be more cautious
        
         | stavros wrote:
         | This makes no sense. It's not like they have a "slow it down"
         | knob, they're probably parallelizing your request so you get a
         | 2.5x speedup at 10x the price.
        
         | Aurornis wrote:
         | > Anthropic offering 2.5x makes me assume they have 5x or 10x
         | themselves.
         | 
         | They said the 2.5X offering is what they've been using
         | internally. Now they're offering via the API:
         | https://x.com/claudeai/status/2020207322124132504
         | 
         | LLM APIs are tuned to handle a lot of parallel requests. In
         | short, the overall token throughput is higher, but the
         | individual requests are processed more slowly.
         | 
         | The scaling curves aren't that extreme, though. I doubt they
         | could tune the knobs to get individual requests coming through
         | at 10X the normal rate.
         | 
         | This likely comes from having some servers tuned for higher
         | individual request throughput, at the expense of overall token
         | throughput. It's possible that it's on some newer generation
         | serving hardware, too.
        
       | jonplackett wrote:
       | Is this is the beginning of the 'Speedy boarding' / 'Fastest
       | delivery' enshitification?
       | 
       | Where everyone is forced to pay for a speed up because the
       | 'normal' service just gets slower and slower.
       | 
       | I hope not. But I fear.
        
         | falloutx wrote:
         | This is to test the room before real enshitification happens.
         | Companies who bought from Anthropic are really in for a ride.
        
       | dmix wrote:
       | I really like Anthropic's web design. This doc site looks like
       | it's using gitbook (or a clone of gitbook) but they make it look
       | so nice.
        
         | treycluff wrote:
         | Looks like mintlify to me. Especially the copy page button.
        
         | falloutx wrote:
         | Its just https://www.mintlify.com/ with barely customized theme
        
       | henning wrote:
       | LLM programming is very easy. First you have to prompt it to not
       | mistakes. Then you have to tell it to go fast. Software
       | engineering is over bro, all humans will be replaced in 6 days
       | bro
        
       | kristianp wrote:
       | This is gold for Anthropic's profitability. The Claude Code
       | addicts can double their spend to plow through tokens because
       | they need to finish something by a deadline. OpenAI will have a
       | similar product within a week but will only charge 3x the normal
       | rate.
       | 
       | This angle might also be NVidias reason for buying Groq. People
       | will pay a premium for faster tokens.
        
       ___________________________________________________________________
       (page generated 2026-02-07 23:00 UTC)